The SAS macros

This page is the repository's own document docs/sas-macros.md, shown as it is, so that it never drifts from the product.

Made by node tools/sas_macros_doc.ts from the headers of sas/macros/ (the macro library 1.8.0); do not edit it. How a macro is written: sas-style.md; what the library computes and how a program reads: sas.md. The same as data for an agent: sas/macros/index.json (and sapling find <macro>, the MCP tool sapling_sas_macros).

Every macro a person calls checks its parameters first and stops with one line that names the parameter and the fix (ERROR: [describe] var=AGEX is not a variable of adsl: give var= a variable of data=), writes a begin and an end line in the log (%let spl_quiet = yes; keeps them out), and keeps its work data sets with debug=yes.

The global macro variables#

Setting up a program#

The statistics macros (one block of a table each)#

Tables and listings#

Figures (one macro per figure: docs/sas.md "Figures")#

The steps of an ADaM program#

The inner parts (a person does not call them)#

%_spl_arg#

One parameter of a call that a one-call table writes into its long program: its name and its value. A parameter without a value is left out.

File: sas/macros/_spl_arg.sas (13NOV2026)

%_spl_arg(data, "&rec.")

Output: CODE: DATA step statements that add the parameter to the arrays PN and PV of %_spl_program(begin).

%_spl_begin#

The begin line of a macro in the log, with the version of the macro library; it gives the time it began, for the end line (%_spl_end). The version below is the one place of the version of the library: every header says the same.

File: sas/macros/_spl_begin.sas (13NOV2026)

%_spl_begin(describe)

Output: TEXT: the text of the time it began (seconds); the line NOTE: [describe] begins (Sapling macro library 1.1.0). in the log, unless SPL_QUIET is yes (a global macro variable: %LET spl_quiet = yes; before the calls keeps the log short).

%_spl_call#

One call of a macro in the long program of a one-call table: the parameters given by %_spl_arg since the last call, one per line, the equal signs under each other.

File: sas/macros/_spl_call.sas (13NOV2026)

%_spl_call(freq)

Output: CODE: DATA step statements that write the call to the file of the long program and start the next list of parameters.

%_spl_chars#

The character variables among some variables of a data set (their names, separated by blanks). The statistics macros give them one length ($200) in their results, so that results stack without a cut value.

File: sas/macros/_spl_chars.sas (12NOV2026)

%_spl_chars(adae, TRT01A AEBODSYS AEDECOD AVISITN)

Output: TEXT: the names of the character variables among VARS, separated by blanks (a variable not in DATA is left out).

%_spl_csv#

Write a data set as a CSV file: a line with the names of the variables, then one line per row (a DATA step; values with a comma or a quote are quoted).

File: sas/macros/_spl_csv.sas (12NOV2026)

%_spl_csv(data=l14_3_1(KEEP=USUBJID AEDECOD), file=&spl_out./listings/l14_3_1.csv)

Output: FILE: a line with the names of the variables, then one line per row.

%_spl_empty#

The result of a statistics macro when no record qualifies: the result shape without a row, and a NOTE that says so. A procedure never runs on no record, so the log has no error of one.

File: sas/macros/_spl_empty.sas (12NOV2026)

%_spl_empty(macro=cox, out=&out., like=_cox_records, keep=&by. &within., add=versus)

Output: OUT: KEEP, ADD, GROUPS, BLOCK, STAT, VALUE and TEXT and no row; the line NOTE: [macro] No data qualified ... in the log.

%_spl_end#

The end line of a macro in the log: its run time since its begin line (%_spl_begin), and with debug=yes that its work data sets are kept.

File: sas/macros/_spl_end.sas (12NOV2026)

%_spl_end(describe, &t0., &debug.)

Output: LOG: NOTE: [describe] ends after 0.42 seconds. in the log, unless SPL_QUIET is yes; with debug=yes always the line that names the kept work data sets.

%_spl_figure_attrs#

The look of the groups of a figure: an attribute map (DATTRMAP=) that gives each value of the group variable a colour, a marker and a line pattern, in the order of the columns of %columns (columns=; a value not in it after them), else in the order the values come in the data. Eight colours safe for colour blindness (Okabe and Ito, without the yellow, and a grey), each with another marker and line pattern, so that groups differ in black and white too; a ninth group starts the list again. ID grp has all attributes; ID grpcol the colours only, solid lines (an error bar; a marker of its own, such as the censoring mark: give it to a copy of the group variable, as SGPLOT takes one ID per variable). The data get _ORDER, the place of their group.

File: sas/macros/_spl_figure_attrs.sas (12NOV2026)

%_spl_figure_attrs(data=_kmplot_plot, columns=&columns., var=&by., out=_kmplot_attrs)

Output: OUT: the attribute map: ID, VALUE, _ORDER and the attributes; DATA: data= with _ORDER.

%_spl_figure_close#

Close the destinations of a figure (%_spl_figure_open) and set back what it changed: the RTF file closed, LISTING closed and opened again (its style LISTING) when it was open before, the options of ODS GRAPHICS reset.

File: sas/macros/_spl_figure_close.sas (12NOV2026)

%_spl_figure_close(listing=&listing.)

Output: CODE: the ODS statements that close the destinations.

%_spl_figure_dodge#

Move the groups a little apart on a numeric x axis, so that their error bars do not overlap: _XPLOT = x + (_ORDER - (G + 1) / 2) * STEP, G the number of groups, STEP the smaller of half the smallest distance of two x values divided by G - 1 and 3% of the range of x. The x of the data stays as it is.

File: sas/macros/_spl_figure_dodge.sas (12NOV2026)

%_spl_figure_dodge(data=_meanplot_plot, x=x)

Output: DATA: data= with _XPLOT.

%_spl_figure_label#

The label of an axis, as a text in quotes: the text given (in quotes, or put in quotes), else the label of the variable, else its name.

File: sas/macros/_spl_figure_label.sas (12NOV2026)

%_spl_figure_label(&data., &time., %SUPERQ(xlabel))

Output: TEXT: the label in double quotes.

%_spl_figure_nodata#

No record qualified for a figure: out= in the shape of the figure with no row, the NOTE No data qualified, and the figure that says so (a blank plot with the text, in the PNG and the RTF of file=, under the titles of the program).

File: sas/macros/_spl_figure_nodata.sas (12NOV2026)

%_spl_figure_nodata(
    keep   = &by.
  , macro  = kmplot
  , like   = _kmplot_records
  , chars  = kind
  , nums   = x y n censored
  , file   = &file.
  , width  = &width.
  , height = &height.
  , out    = &out.
)

Output: OUT: no row; FILE: <file>.png and <file>.rtf with the text.

%_spl_figure_open#

Open the destinations of one figure: the PNG image (the LISTING destination with GPATH=, 300 dpi, image name and size from ODS GRAPHICS) and the RTF file with the figure on a page (the titles and footnotes of the program as text of the page: NOGTITLE NOGFOOTNOTE). The PNG carries the titles and the footnotes in the image (LISTING draws them in the graph, as SAS does by default), so a PNG alone says what it is. Both use the style HTMLBLUE: tick values 9pt, axis labels and legend 10pt, titles 11pt. It notes in the macro variable named by state= whether LISTING was open, for %_spl_figure_close.

File: sas/macros/_spl_figure_open.sas (12NOV2026)

%_spl_figure_open(file=&file., width=&width., height=&height., state=listing)

Output: CODE: the ODS statements that open the destinations.

%_spl_figure_outpar#

The drawing parameters of a figure macro as a data set (its outpar=): one row per parameter in the order of the signature, PARAMETER its name in lower case and VALUE its value as the macro draws it (given, from par=, the default, or made by the macro: the label of an axis, the format of the ticks). par= of a later call reads it back.

File: sas/macros/_spl_figure_outpar.sas (13NOV2026)

%_spl_figure_outpar(names=&drawn., out=&outpar.)

Output: DATA: out= with PARAMETER ($32) and VALUE ($500).

%_spl_figure_par#

The parameters of a figure macro, before its checks. With plot= (draw the out= of a call again): data= is not given, plot= has the variables of out=, and each parameter of the computation that has a value gets a NOTE (it is not used). With par= (the outpar= of a call): each drawing parameter that the call leaves blank gets the VALUE of its row; a row that names no drawing parameter is an error. Then each parameter still blank gets its default from draw= or compute=. So a value in the call wins over a row of par=, and a row of par= over the default: the defaults of these parameters are here, not in the signature of the macro.

File: sas/macros/_spl_figure_par.sas (13NOV2026)

%_spl_figure_par(
    macro   = kmplot
  , draw    = columns by panel atrisk=yes xlabel ylabel="Survival (%)" file
  , compute = where time cnsr=CNSR
  , shape   = kind x y lower upper n censored
  , names   = drawn
  , state   = ok
)

Output: LOG: one ERROR with its fix for a wrong plot= or par=, a NOTE for each parameter not used. The local macro variables of the caller: the parameters set from par= and from the defaults, state= and names=.

%_spl_levels#

The planned categories of some variables as rows: every combination of their values, by DO loops. A character variable gets the length $200.

File: sas/macros/_spl_levels.sas (12NOV2026)

%_spl_levels(levels=BNRIND 'LOW' 'NORMAL' 'HIGH' ANRIND 'LOW' 'NORMAL' 'HIGH', data=adlbc)

Output: CODE: DATA step statements: LENGTH of the character variables ($200), then DO loops that output every combination of the planned values (DO BNRIND = ...; DO ANRIND = ...; OUTPUT; END; END;).

%_spl_need#

Check the parameters a macro needs: each must have a value, and a data set of data=, denom=, columns= or from= must exist. For each one that does not it writes one line that names the parameter and the fix: ERROR: [describe] data= is required: give data=<value> ... ERROR: [describe] data=adsx does not exist: give data= ... It gives 1 when all are there, 0 when one is not: the caller then stops (%RETURN).

File: sas/macros/_spl_need.sas (13NOV2026)

%_spl_need(describe, data var by out)

Output: TEXT: 1 or 0, as text.

%_spl_nobs#

The number of rows of a data set (with the WHERE= of its data set options, when it has one); 0 when the data set does not exist.

File: sas/macros/_spl_nobs.sas (12NOV2026)

%_spl_nobs(_describe_values)

Output: TEXT: the number, as text.

%_spl_oneof#

Check that the value of a parameter is one of a list (in any case). When it is not it writes one line that names the parameter and the fix: ERROR: [cox] ties=fast is not one of efron breslow exact ... It gives 1 or 0: the caller then stops (%RETURN).

File: sas/macros/_spl_oneof.sas (12NOV2026)

%_spl_oneof(cox, ties, efron breslow exact discrete)

Output: TEXT: 1 or 0, as text.

%_spl_pairlist#

The pairs of groups of a parameter as rows: _K (the place of the pair), _A (the group) and _B (the group it is compared with). The pairs are separated by #, one pair per line, the first group against the second; a value with a blank or a hyphen is in quotes: pairs = "Xanomeline High Dose" "Placebo" # "Xanomeline Low Dose" "Placebo" The form of version 1.0 is still read: "A"-"B" "C"-"D" (no #, every item a pair joined by a hyphen outside quotes).

File: sas/macros/_spl_pairlist.sas (12NOV2026)

%_spl_pairlist(list=pairs)

Output: CODE: DATA step statements: one row (OUTPUT) per pair with _K, _A and _B ($200).

%_spl_pairs#

The differences of the pairs of groups asked for, from the table Diffs of PROC MIXED (it has each pair once, in one direction). It computes nothing: a difference the other way round is the same difference with its sign and its limits turned.

File: sas/macros/_spl_pairs.sas (12NOV2026)

%_spl_pairs(
    by      = TRT01P
  , within  =
  , diffs   = _lsmeans_diffs
  , records = _lsmeans_records
  , pairs   = all
  , order   = TRT01PN
  , out     = _lsmeans_pairs
)

Output: OUT: WITHIN, BY (the group), VERSUS (the group it is compared with), diff, diff_se, lcl, ucl, pval: one row per pair asked for (in each group of WITHIN).

%_spl_program#

The fixed parts of the DATA step with which a one-call table writes its long program: begin (the file and the variables), records (the population and the records, from the parameters ADSL, POP, DATA, WHERE and NAME of the calling macro), table (the %table call, from BY, ORDER, TOTAL, FORMAT, FILE, NAME and DEBUG), end (the routine that writes a call). It computes nothing.

File: sas/macros/_spl_program.sas (13NOV2026)

%_spl_program(begin)

Output: CODE: DATA step statements. It reads the macro variables REC (the records the calls read), POPN (the data set with the N of the columns), DS and NAMES of the calling macro.

%_spl_quote#

The value of a parameter as a text in single quotes, for a DATA step: each single quote in it doubled, nothing in it resolved. The code the macro writes holds the value itself (SYMGET would read it when the step runs), so that the flat program of sapling sas run --flat (the code without a macro) has it too. An empty value gives a blank text.

File: sas/macros/_spl_quote.sas (12NOV2026)

%_spl_quote(label)

Output: TEXT: a text in single quotes: 'Age (years)', '"A" "B" # "C" "B"'.

%_spl_rows#

The result shape: one row per statistic of the row just read. Each statistic is a numeric variable of the same name (n, mean, pval ...); the row gets STAT (its name), VALUE (its number), PATTERN and TEXT (the number as printed, %_spl_text: a p value, pval or padj, never as 0 or 1).

File: sas/macros/_spl_rows.sas (13NOV2026)

%_spl_rows(stats=n="XX" mean="XX.X" sd)

Output: CODE: DATA step statements: one row per statistic with STAT, VALUE and TEXT. The default patterns: XX for n, events, min, max; XX.X for mean, median, q1, q3, lsmean, diff, estimate, lcl, ucl; (XX.X%) for pct; X.XXX for pval; XX.XX for sd, se, lsmean_se, diff_se and any other statistic.

%_spl_stackpairs#

The records of each pair of groups asked for, one under the other: the pair (_PAIR), its group (_A) and the group it is compared with (_B), and the side of each record (_SIDE: 1 in _A, 2 in _B). A procedure then runs BY _PAIR _A _B on the 2 groups of each pair, the group first. It computes nothing.

File: sas/macros/_spl_stackpairs.sas (13NOV2026)

%_spl_stackpairs(
    macro   = ttest
  , records = _ttest_records
  , by      = TRT01P
  , pairs   = all
  , order   = TRT01PN
  , out     = _ttest_pairs
)

Output: OUT: the records of each pair with _PAIR, _A, _B and _SIDE (1 in _A, 2 in _B), sorted by them. The macro variable PAIRSOK of the caller: 1, or 0 when a pair names a group that no record has (an error in the log).

%_spl_text#

The text of a number by its pattern. A pattern shows one number as one group of X: XX, XX.X, (XX.X%), (N=XX). The number is rounded (ROUND) to the decimals of the X group, right aligned in its width and put in its place; the text is trimmed. A p value never prints as 0 or 1: rounded to 0 it is < and the smallest step (X.XXX: <0.001), rounded to 1 it is > and one step less (>0.999). The number is unchanged.

File: sas/macros/_spl_text.sas (12NOV2026)

%_spl_text(
      value   = VALUE
    , pattern = PATTERN
    , text    = TEXT
    , pvalue  = STAT IN ('pval' 'padj')
  )

Output: CODE: DATA step statements: TEXT is the number of VALUE as its PATTERN shows it (9.46 with (XX.X%) gives ( 9.5%); 86 with (N=XX) gives (N=86); 3 with XX gives 3; a p value 0.00004 with X.XXX gives <0.001).

%_spl_total#

The name of the group of all groups together, from total=: nothing for no, Total for yes, else the text in quotes.

File: sas/macros/_spl_total.sas (12NOV2026)

%_spl_total(%SUPERQ(total))

Output: TEXT: nothing for no (or empty), Total for yes, else the text of the quotes.

%_spl_vars#

Check that the variables of some parameters are variables of a data set. For each one that is not it writes one line that names the parameter and the fix: ERROR: [describe] var=AGEX is not a variable of adsl: give ... It gives 1 when all are there, 0 when one is not: the caller then stops (%RETURN). A data set that cannot be opened is the business of %_spl_need.

File: sas/macros/_spl_vars.sas (12NOV2026)

%_spl_vars(describe, data, var by)

Output: TEXT: 1 or 0, as text.

%_spl_windows#

The rows of windows= (or visits=) of a call as a data set: one row per window, the rows of the call separated by #, the values of a row by blanks, a text in quotes, a missing number as a period.

File: sas/macros/_spl_windows.sas (13NOV2026)

%_spl_windows(windows, windows, day, _windows_rows)

Output: DATA: _out, one row per window in the order of the call. The macro variable ROWSOK of the caller: 1, or 0 when a row has not its number of values (an error in the log).

%adjust#

The adjusted p-values of a family of pairs of groups (the hypotheses of a multiplicity block): the p-values (STAT = pval) of a result of pairs (%lsmeans, %cox, %logistic, %risk, %rateratio, %ttest, %wilcoxon), in the order of family=, adjusted by method= (PROC MULTTEST INPVALUES=: bonferroni, holm (STEPBON), hochberg; a DATA step: bonferroni with weights, hierarchy, the fixed sequence: the largest p-value of a hypothesis and those before it). The result is the input with a row padj for each pair of the family.

Procedure: PROC MULTTEST (INPVALUES=); a DATA step
Proven: the known answer of the method riskdiff (src/methods.ts, worked out by tools/method_answers.R; node tools/methods_sas.ts) with the adjusted p-values of a family; ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/adjust.sas (12NOV2026)

%adjust(
    data   = resp_rd
  , by     = TRT01P
  , family = "High" "Placebo"
           # "Low"  "Placebo"
  , method = hochberg
  , out    = resp_rd
)

Output: OUT: the rows of data= and a row padj per pair of the family.

%aeplot#

The adverse events of the subjects as two panels side by side that share the axis of the terms: at the left the percent of the subjects of each group with the term (dots), at the right the risk difference of each group against the reference group with its interval and a line at 0. The flags are those of %termflags, the risk differences those of %risk (ci=, level=), so that the figure and a table made with them cannot differ. The terms: those of threshold=, sorted by sort=, at most max= of them.

Procedure: PROC SUMMARY, PROC SGPANEL (and PROC FREQ in %risk)
Proven: its known answer in tools/figures_sas.ts (percents and Wald intervals worked out by hand); not yet run on a SAS server
File: sas/macros/aeplot.sas (13NOV2026)

%aeplot(
    data      = adae
  , columns   = t14_5_1_cols
  , denom     = adsl
  , where     = (TRTEMFL = 'Y')
  , term      = AEDECOD
  , by        = TRTA
  , reference = "Placebo"
  , threshold = 5
  , file      = &spl_out./figures/f14_5_3
  , out       = f14_5_3
)

Output: OUT: TERM and BY, KIND, X, LOWER, UPPER, N. KIND percent: one row per term and group, X the percent of its subjects, N the subjects with the term. KIND difference: one row per term and group that is not the reference, X the risk difference in percent (100 x RD of %risk), LOWER and UPPER its limits in percent. OUTPAR: one row per drawing parameter in the order of the table: PARAMETER ($32) its name, VALUE ($500) its value as drawn (a label in quotes). FILE: <file>.png and <file>.rtf.

%ancova#

A linear model of a variable, one record per subject (PROC GLM): class factors, covariates and the tested term. The Type III test of the tested term (pval); its coefficient, its standard error and confidence limits (estimate, se, lcl, ucl) when stats= asks for them.

Procedure: PROC GLM
Proven: the known answer of the method ancova (src/methods.ts, worked out by tools/method_answers.R; node tools/methods_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/ancova.sas (12NOV2026)

%ancova(
    data   = advs
  , where  = (PARAMCD = 'SYSBP' and AVISIT = 'Week 12')
  , var    = CHG
  , covars = BASE
  , class  = SITEGR1
  , test   = TRTPN
  , out    = chg_test
)

Output: OUT: one row per statistic, of all subjects together (GROUPS is blank).

%anova#

The one-way analysis of variance of a variable between groups (PROC GLM): the F test of the model (pval).

Procedure: PROC GLM
Proven: the known answer of the method anova (src/methods.ts, worked out by tools/method_answers.R; node tools/methods_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/anova.sas (12NOV2026)

%anova(
    data  = adsl
  , var   = AGE
  , by    = TRT01A
  , label = "Age (years)"
  , out   = age_test
)

Output: OUT: one row, of all groups together (GROUPS is blank).

%ard#

Every number of a table as a CSV file, with every digit: one line per number with the result it comes from, its statistic and the values of its groups (in the order of GROUPS), then the number and its text. The N of each column comes first (result N). It reads the results, never the printed page: sapling sas check compares these numbers with those of R.

Procedure: DATA steps (FILE, PUT)
Proven: the tables of the dummy study SYNHTN01 (sapling sas check: SAS CHECK PASS, 759 of 759 numbers); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/ard.sas (12NOV2026)

%ard(data=t14_2_1, columns=t14_2_1_cols)

Output: FILE: <out>/ard/<file>.csv: result, stat, g1_value ... g5_value, raw, formatted; the N of each column first (result N), then every number of the results.

%barplot#

Bars of the categories of a variable in each group: the percent of the subjects of the group (denom=: the subjects of the population; a subject counts once in a category) or of its records, or the count; the groups side by side or stacked, the value on each bar. One panel per value of panel=.

Procedure: PROC FREQ, PROC SGPANEL
Proven: its known answer in tools/figures_sas.ts (counts and percents worked out by hand); not yet run on a SAS server
File: sas/macros/barplot.sas (13NOV2026)

%barplot(
    data    = adae
  , columns = t14_5_1_cols
  , denom   = adsl
  , where   = (TRTEMFL = 'Y')
  , var     = AESEV
  , by      = TRTA
  , file    = &spl_out./figures/f14_5_2
  , out     = f14_5_2
)

Output: OUT: one row per bar: PANEL and BY, KIND bar, X (the category as printed), Y (the percent or the count), N (the subjects or records of the bar), DENOM (the N of the percent). A category without a record in a group has no bar. OUTPAR: one row per drawing parameter in the order of the table: PARAMETER ($32) its name, VALUE ($500) its value as drawn (a label in quotes). FILE: <file>.png and <file>.rtf.

%baseline#

Give every record of a group the value of the group's baseline record (the one with the baseline flag).

Procedure: PROC SORT, MERGE
Proven: the fixtures of the step kind baseline (src/adam_steps.ts; node tools/adam_steps_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/baseline.sas (12NOV2026)

%baseline(
    data  = advs
  , value = AVAL
  , by    = USUBJID PARAMCD
  , new   = BASE
)

Output: DATA: NEW on every record: the VALUE of the baseline record of its group (missing for a group without one); two baseline records in a group are an ERROR that names the group. The rows keep their order.

%boxplot#

Box plots of a variable in each group (and at each visit of x=): the box from Q1 to Q3 with the median, the whiskers to the most extreme values within 1.5 times the interquartile range of the box, the values beyond as outliers, the mean as a marker. The quantiles are those of PROC MEANS QNTLDEF=5 and of VBOX (its default PERCENTILE=5; R quantile type 2). One panel per value of panel=.

Procedure: PROC MEANS, PROC SGPANEL
Proven: its known answer in tools/figures_sas.ts (quartiles and whiskers worked out by hand); not yet run on a SAS server
File: sas/macros/boxplot.sas (13NOV2026)

%boxplot(
    data    = advs
  , columns = t14_3_1_cols
  , where   = (PARAMCD = 'SYSBP' and ANL01FL = 'Y')
  , y       = CHG
  , x       = AVISITN
  , by      = TRTA
  , xformat = VISIT.
  , ref     = 0
  , file    = &spl_out./figures/f14_3_1
  , out     = f14_3_1
)

Output: OUT: one row per box: PANEL, BY, KIND box, X (the category; missing without x=), N, MEAN, MEDIAN, Q1, Q3, MIN, MAX, WLOW and WHIGH (the ends of the whiskers: the smallest and the largest value within Q1 - 1.5 IQR and Q3 + 1.5 IQR). OUTPAR: one row per drawing parameter in the order of the table: PARAMETER ($32) its name, VALUE ($500) its value as drawn (a label in quotes). FILE: <file>.png and <file>.rtf.

%chisq#

The Pearson chi-square test (without continuity correction) of groups against categories, each subject once in each group and category (PROC FREQ): pval. With exact=yes Fisher's exact test of the same table.

Procedure: PROC FREQ (CHISQ, FISHER)
Proven: the known answer of the methods chisq, fisher (src/methods.ts, worked out by tools/method_answers.R; node tools/methods_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/chisq.sas (12NOV2026)

%chisq(
    data  = adsl
  , var   = SEX
  , by    = TRT01A
  , label = "Sex"
  , out   = sex_test
)

Output: OUT: one row, or one per group of WITHIN (GROUPS: their names). Where PROC FREQ makes no test (a table with one row or one column) the value is missing.

%cmh#

The Cochran-Mantel-Haenszel test of groups against categories in strata (PROC FREQ), each subject once in each group, category and stratum: pval. The test of general association, or (statistic=rowmean) the test that the mean scores of the groups differ, the scores being the values of the numeric response.

Procedure: PROC FREQ (CMH)
Proven: the known answer of the method cmh (src/methods.ts, worked out by tools/method_answers.R; node tools/methods_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/cmh.sas (12NOV2026)

%cmh(
    data      = adqscibc
  , var       = AVAL
  , by        = TRT01P
  , strata    = SITEGR1
  , within    = AVISIT
  , statistic = rowmean
  , out       = cibic_test
)

Output: OUT: one row, or one per group of WITHIN (GROUPS: their names). Where PROC FREQ makes no test the value is missing.

%columns#

The columns of a table: each value of the column variable in the results (and in the population), in the order of ORDER (else of the values), the total last; the N of each column from the population (its subjects, the total all of them) and the header of the column with its N: Placebo|(N=86).

Procedure: PROC SUMMARY
Proven: the tables of the dummy study SYNHTN01 (sapling sas check: the N of every column the same as R); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/columns.sas (12NOV2026)

%columns(
    data  = t14_2_1
  , n     = adsl
  , var   = trt01a
  , order = trt01an
  , total = yes
  , out   = t14_2_1_cols
)

Output: OUT: one row per column: COL (1, 2, ...), COLVALUE (the value), BIGN (the N) and HEADER (the value, a | and (N=86): the split character of %print).

%cox#

A Cox proportional hazards model of a time, one record per subject (PROC PHREG): the groups, the covariates and the class factors, stratified by the strata (each combination of their values a stratum with its own baseline hazard). The hazard ratio of pairs of groups from one fit (hr), its Wald confidence limits (lcl, ucl) and the Wald p-value of the pair (pval). With pvalue=logrank the p-value is the log-rank test of the two groups of the pair, stratified by the same strata (PROC LIFETEST); the hazard ratio and its limits stay those of the model.

Procedure: PROC PHREG (LSMEANS); PROC LIFETEST for pvalue=logrank
Proven: the known answer of the method cox (src/methods.ts, worked out by tools/method_answers.R; node tools/methods_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/cox.sas (12NOV2026)

%cox(
    data   = adtte
  , where  = (PARAMCD = 'PFS')
  , var    = AVAL
  , by     = TRT01P
  , strata = STRAT1 STRAT2
  , order  = TRT01PN
  , out    = pfs_hr
)

Output: OUT: one row per pair and statistic (BY: the group, VERSUS: the group it is compared with); GROUPS is BY VERSUS (and WITHIN).

%csv#

The records of a listing (or the numbers of a figure, out= of a figure macro) as a CSV file: a line with the names of the variables, then one line per record in the order of the data. sapling sas check compares it with the listing or the figure of R.

Procedure: DATA steps (FILE, PUT)
Proven: the listings of the dummy study SYNHTN01 (sapling sas check: listings the same in R and SAS); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/csv.sas (12NOV2026)

%csv(data=adae, var=trt01a usubjid aedecod astdt, file=l14_5_1)

Output: FILE: <out>/<folder>/<file>.csv: the names of VAR, then one line per record.

%describe#

Descriptive statistics of a variable in each group: n, mean, SD, median, minimum and maximum of the values that are not missing; Q1, Q3 and the standard error when stats= asks for them.

Procedure: PROC MEANS
Proven: the known answer of the method contsumm (src/methods.ts, worked out by tools/method_answers.R); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/describe.sas (12NOV2026)

%describe(
    data  = adsl
  , var   = AGE
  , by    = TRT01A
  , total = yes
  , label = "Age (years)"
  , out   = age
)

Output: OUT: one row per group and statistic: the BY variables (and KEEP), GROUPS (the names of the BY variables), BLOCK (the label), STAT, VALUE (the number) and TEXT (the number as printed). A group without a value has no row; no value at all: no row and the NOTE "No data qualified".

%fisher#

Fisher's exact test of "subjects with a record against subjects without" between groups (PROC FREQ): pval. Every subject of DENOM is counted, with a record or without: the usual test of an adverse event between treatments. With exact=no the Pearson chi-square test of the same table.

Procedure: PROC FREQ (FISHER, CHISQ)
Proven: the known answer of the method fisher (src/methods.ts, worked out by tools/method_answers.R; node tools/methods_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/fisher.sas (12NOV2026)

%fisher(
    data   = adae
  , denom  = adsl
  , where  = (TRTEMFL = 'Y')
  , by     = TRT01A
  , within = AEBODSYS AEDECOD
  , out    = pt_test
)

Output: OUT: one row, or one per group of WITHIN (GROUPS: their names). Where PROC FREQ makes no test the value is missing.

%flag_first#

Flag one record of each group: the first (or last) one, among the records that meet a condition. The other records keep a blank flag.

Procedure: PROC SORT, MERGE
Proven: the fixtures of the step kind flag.extreme (src/adam_steps.ts; node tools/adam_steps_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/flag_first.sas (13NOV2026)

%flag_first(
    data  = advs
  , where = (not missing(AVISITN) and not missing(AVAL))
  , by    = USUBJID PARAMCD AVISITN
  , order = AWTDIFF ADY VSSEQ
  , new   = ANL01FL
)

Output: DATA: NEW: Y on the first (last) record of each group among the records that meet where=; the others keep a blank NEW. The rows keep their order.

%flag_if_any#

Flag the subjects that have a record in another data set that meets a condition.

Procedure: PROC SORT, MERGE
Proven: the fixtures of the step kinds flag.condition, flag.any (src/adam_steps.ts; node tools/adam_steps_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/flag_if_any.sas (12NOV2026)

%flag_if_any(
    data  = adsl
  , from  = sdtm.ds
  , where = (DSDECOD = 'COMPLETED')
  , new   = COMPLFL
)

Output: DATA: NEW: Y on every row of a subject that has such a record in FROM, else N (or blank with else=blank), in the order of its rows.

%forestplot#

The forest plot of results in the levels of subgroups: one row per level under a heading row of its factor, the estimate as a marker with its interval as a bar, a line at no difference, a log axis for a ratio, and the estimate (lower, upper) as text at the right. It draws results (of %cox, %logistic, %rateratio, %lsmeans or %risk with within=), so that the figure and the table of the same results cannot differ.

Procedure: PROC SUMMARY, PROC SGPLOT
Proven: its known answer in tools/figures_sas.ts (made-up results, arranged by hand); not yet run on a SAS server
File: sas/macros/forestplot.sas (13NOV2026)

%forestplot(
    data     = t14_2_sg
  , estimate = hr
  , within   = SEX AGEGR1 RACE
  , by       = TRTA
  , scale    = log
  , ref      = 1
  , file     = &spl_out./figures/f14_2_sg
  , out      = f14_2_sg
)

Output: OUT: one row per level and pair: FACTOR (the variable of within=), LEVEL (its value as printed), PAIR (by= vs VERSUS; blank without by=), KIND estimate, X (the estimate), LOWER, UPPER, N (missing unless the results have n). OUTPAR: one row per drawing parameter in the order of the table: PARAMETER ($32) its name, VALUE ($500) its value as drawn (a label in quotes). FILE: <file>.png and <file>.rtf.

%freq#

Frequencies of subjects (PROC FREQ): the subjects of each category in each group (n), their percent of the subjects of the group (pct), and the records (events) when stats= asks for them. A subject is counted once in each cell. Without var= the cell is "a record at all".

Procedure: PROC FREQ
Proven: the known answer of the methods count, catsumm (src/methods.ts, worked out by tools/method_answers.R; node tools/methods_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/freq.sas (13NOV2026)

%freq(
    data  = adsl
  , var   = SEX
  , by    = TRT01A
  , total = yes
  , label = "Sex"
  , out   = sex
)

Output: OUT: one row per group, category and statistic: BY and VAR (and KEEP), GROUPS (their names), BLOCK, STAT, VALUE, TEXT. Every group of DENOM has every category: n = 0 when no subject has it.

%import_xpt#

Read SAS transport files (XPORT version 5) into SAS data sets.

Procedure: LIBNAME XPORT, PROC COPY
Proven: the ADaM and SDTM transport files (sapling sas run, sapling sas adam run) of the dummy study SYNHTN01 (sapling sas run: SAS RUN PASS; sapling sas check: SAS CHECK PASS, 570 of 570 numbers of R the same in SAS); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/import_xpt.sas (12NOV2026)

%import_xpt(
    dir     = /data/htn3/xpt
  , members = adsl adae advs
  , out     = /data/htn3/adam
)

Output: FILE: the data sets <out>/<member>.sas7bdat. The macro stands alone: a program includes it without the library (00_import.sas), so it calls no other macro of the library and writes no begin and end lines.

%interaction_genmod#

The test of the interaction of the treatment and a factor in a count model with its exposure, one record per subject (PROC GENMOD, the logarithm of the exposure the offset): Poisson or negative binomial (the dispersion k estimated in each model), the model of the groups, the covariates, the class factors and the factor, against the same with the treatment-by-factor terms. The likelihood-ratio test (pval): twice the difference of the full log likelihoods of the two fits, with the difference of the numbers of their estimated coefficients.

Procedure: PROC GENMOD
Proven: the known answer of the methods interaction-poisson, interaction-negbin (src/methods.ts, worked out by tools/method_answers.R; node tools/methods_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/interaction_genmod.sas (12NOV2026)

%interaction_genmod(
    data     = adsl
  , var      = EXACNT
  , by       = TRT01P
  , factor   = SEX
  , exposure = TRTDUR / 365.25
  , dist     = negbin
  , label    = "Sex"
  , out      = sex_int
)

Output: OUT: one row, of all groups and levels together (GROUPS is blank).

%interaction_glm#

The test of the interaction of the treatment and a factor in a linear model, one record per subject (PROC GLM): the variable on the groups, the covariates, the class factors, the factor and the treatment-by-factor terms; the Type III F test of the treatment-by-factor terms (pval). For the term of the highest order this is the F test of the model with the factor against the same model with the terms.

Procedure: PROC GLM
Proven: the known answer of the method interaction-lm (src/methods.ts, worked out by tools/method_answers.R; node tools/methods_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/interaction_glm.sas (12NOV2026)

%interaction_glm(
    data   = adqsadas
  , where  = (PARAMCD = 'ACTOT' and AVISIT = 'Week 24')
  , var    = CHG
  , by     = TRTP
  , factor = SEX
  , covars = BASE
  , class  = SITEGR1
  , label  = "Sex"
  , out    = sex_int
)

Output: OUT: one row, of all groups and levels together (GROUPS is blank).

%interaction_logistic#

The test of the interaction of the treatment and a factor in a logistic regression, one record per subject (PROC LOGISTIC, the event EVENT=): the model of the groups, the covariates, the class factors and the factor, against the same with the treatment-by-factor terms. The likelihood-ratio test (pval): the difference of the likelihood-ratio chi-squares of the two fits, with the difference of their degrees of freedom.

Procedure: PROC LOGISTIC
Proven: the known answer of the method interaction-logistic (src/methods.ts, worked out by tools/method_answers.R; node tools/methods_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/interaction_logistic.sas (12NOV2026)

%interaction_logistic(
    data   = adsl
  , var    = RESPFL
  , event  = "Y"
  , by     = TRT01A
  , factor = SEX
  , covars = AGE
  , label  = "Sex"
  , out    = sex_int
)

Output: OUT: one row, of all groups and levels together (GROUPS is blank).

%interaction_phreg#

The test of the interaction of the treatment and a factor in a Cox model, one record per subject (PROC PHREG): the model of the groups, the covariates, the class factors and the factor, stratified by the strata, against the same with the treatment-by-factor terms. The likelihood-ratio test (pval): the difference of the likelihood-ratio chi-squares of the two fits (twice the difference of the log partial likelihoods), with the difference of their degrees of freedom.

Procedure: PROC PHREG
Proven: the known answer of the method interaction-cox (src/methods.ts, worked out by tools/method_answers.R; node tools/methods_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/interaction_phreg.sas (12NOV2026)

%interaction_phreg(
    data   = adtte
  , where  = (PARAMCD = 'PFS')
  , var    = AVAL
  , by     = TRT01P
  , factor = SEX
  , strata = STRAT1
  , label  = "Sex"
  , out    = sex_int
)

Output: OUT: one row, of all groups and levels together (GROUPS is blank).

%km#

The Kaplan-Meier summary of a time in each group (PROC LIFETEST): the subjects (n), the events (events), the median time (median) and its confidence limits (lcl, ucl; the limits of Brookmeyer and Crowley). A median the curve does not reach is missing. With time= the survival at that time in percent (surv; TIMELIST=, OUTSURV=) and its limits (lcl, ucl: then of the survival, not of the median).

Procedure: PROC LIFETEST
Proven: the known answer of the methods km, kmrate (src/methods.ts, worked out by tools/method_answers.R; node tools/methods_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/km.sas (13NOV2026)

%km(
    data  = adtte
  , where = (PARAMCD = 'TTDE')
  , var   = AVAL
  , by    = TRT01A
  , out   = ttde
)

Output: OUT: one row per group and statistic.

%kmplot#

The Kaplan-Meier curve of a time in each group: the survival in percent as steps, a mark at each censored time, the number at risk under the axis at its ticks (the subjects whose time is that time or later), in the order of the groups; with median=yes the median time of each group and its interval.

Procedure: PROC LIFETEST, PROC SUMMARY, PROC SGPANEL
Proven: its known answer in tools/figures_sas.ts (survival and number at risk worked out by hand); not yet run on a SAS server
File: sas/macros/kmplot.sas (13NOV2026)

%kmplot(
    data    = adtte
  , columns = t14_1_cols
  , where   = (PARAMCD = 'TTDE')
  , time    = AVAL
  , cnsr    = CNSR
  , by      = TRTA
  , median  = yes
  , xlabel  = "Days to the first dermatologic event"
  , file    = &spl_out./figures/f14_1
  , out     = f14_1
)

Output: OUT: one row per number drawn: PANEL and BY, KIND, X, Y, LOWER, UPPER, N, CENSORED. KIND curve: X a time (0 and each time with an event or a censoring), Y the survival in percent after it, CENSORED 1 when a subject is censored at it. KIND atrisk: X a tick, N the number at risk. KIND median: X the median, LOWER and UPPER its limits (missing: not reached). OUTPAR: one row per drawing parameter in the order of the table, PARAMETER ($32) its name and VALUE ($500) its value as drawn (a label in quotes). FILE: <file>.png and <file>.rtf.

%layout#

The display data set of a table: one row per printed line, as the shells have it. The label of a block, the levels of its groups indented by two blanks each (a term under its class, a visit under its parameter, vs B under A), the statistics of a line put in the template of the line; one column per group with its header and N from %columns, and the column All groups for a p-value of all groups; the p-value of each category (a test of each term, a block of the same label) stands in the row of its category. A category that format= names shows its label. It computes nothing: every number is the TEXT of a result. The line table (Step 3), line: label = template: 0 a category of a frequency block = {n} {pct} {events}; 1 n = {n}; 2 Events = {events}; 2.5 Patient-years = {exposure}; 3 Mean (SD) = {mean} ({sd}) [{se}] (se only: Mean (SE); sd and se: Mean (SD) [SE]); 4 Median = {median}; 5 Q1, Q3 = {q1}, {q3}; 6 Min, Max = {min}, {max}; 7 LS mean (SE) = {lsmean} ({lsmean_se}); 8 Difference (SE) = {diff} ({diff_se}); 9 the estimate by its STAT (hr Hazard ratio, or Odds ratio, irr Rate ratio, rd Risk difference, rr Relative risk, rate Rate per 100 patient-years, else Estimate) = {estimate}; 10 Confidence interval = ({lcl}, {ucl}); 11 p-value = {pval}; 11.5 Adjusted p-value = {padj}; 12 any other STAT, labelled by it = {value}. A part that is missing leaves out its piece of the template with its brackets or its comma (a confidence interval without its upper limit prints the lower limit alone); a TEXT that has the brackets of its piece already is not bracketed twice; the counts of a frequency block are right aligned to the widest count. With setup= it also keeps the %print call of OUT in a global macro variable of that name: %print(&name.) prints the display data set again after a change. The setup holds data=, by=, var=, together= and label= (row=" " and the labels of label=), and rows=, file= and nodata= when they are given here. A parameter that is not in the setup is added after it: %print(&name., file=...); one that is in it cannot be given twice (SAS stops a call with a keyword given twice): give it to %layout again, or write the %print call.

Procedure: DATA steps, PROC FORMAT, PROC SORT, PROC TRANSPOSE
Proven: the tables of the dummy study SYNHTN01 (sapling sas run: SAS RUN PASS, flat programs 12 of 12) and the cases of setup= in node tools/print_sas.ts; ran on SAS 9.4 M9, 2026-11-12; the 12 tables of the CDISC pilot (the p value of each term in T14_5_2), 2026-11-13
File: sas/macros/layout.sas (13NOV2026)

%layout(
    data    = t14_2_1
  , columns = t14_2_1_cols
  , var     = trt01a
  , format  = sex $sexf.
  , file    = &spl_out./tlf/t14_2_1
  , setup   = t14_2_1_print
  , out     = t14_2_1_display
)

Output: OUT: one row per printed line: GRP (the rows that stay on one page: a block, or a level of its first group), ROW (the label, indented, format $CHARw.), C1 ... Cn (the cells, labelled with the header of their column) and C99 (All groups: only when a result has a p-value of all groups). SETUP: with setup=, the global macro variable of that name: the parameters of %print for OUT, such as data=t14_2_1_display, by=row, var=c:, together=grp, label=row=" " (a NOTE in the log shows it); print with %print(&name.).

%lineplot#

Lines through values given (the LS means of %lsmeans or %mmrm by visit, a rate by month): one line per group, with the interval of lower= and upper= as error bars (the groups a little apart on the x axis) or as a band. Nothing is computed: the figure draws the values of data=. One panel per value of panel=.

Procedure: PROC SGPANEL
Proven: its known answer in tools/figures_sas.ts (the values given); not yet run on a SAS server
File: sas/macros/lineplot.sas (13NOV2026)

%lineplot(
    data    = lsm_chg
  , columns = t14_3_1_cols
  , x       = AVISITN
  , y       = ESTIMATE
  , by      = TRTP
  , lower   = LCL
  , upper   = UCL
  , xformat = VISIT.
  , ylabel  = "LS mean change from baseline (95% CI)"
  , ref     = 0
  , file    = &spl_out./figures/f14_3_2
  , out     = f14_3_2
)

Output: OUT: one row per panel, group and x: PANEL and BY, KIND line, X, Y, LOWER, UPPER. OUTPAR: one row per drawing parameter in the order of the table: PARAMETER ($32) its name, VALUE ($500) its value as drawn (a label in quotes). FILE: <file>.png and <file>.rtf.

%locf#

Last observation carried forward into the analysis windows, as new records (ADaM: DTYPE = 'LOCF'). For each window after the first one after baseline in which a group has no analysed record: a copy of the last analysed record before it, moved to that window. Baseline is not carried.

Procedure: PROC SORT, a DATA step (SET POINT=)
Proven: the fixtures of the step kind impute (src/adam_steps.ts; node tools/adam_steps_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/locf.sas (13NOV2026)

%locf(
    data    = adqs
  , windows = &advs_windows.
  , where   = (PARAMCD = 'ACTOT')
)

Output: DATA: sorted by the groups, AVISITN and DTYPE; _ROW is the new order.

%logistic#

A logistic regression of a response with two values, one record per subject (PROC LOGISTIC): the probability of the event (event=) on the groups, the covariates and the class factors. The odds ratio of pairs of groups from one fit (or), its Wald confidence limits (lcl, ucl) and the Wald p-value of the pair (pval).

Procedure: PROC LOGISTIC (LSMEANS)
Proven: the known answer of the method logistic (src/methods.ts, worked out by tools/method_answers.R; node tools/methods_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/logistic.sas (12NOV2026)

%logistic(
    data   = adrs
  , where  = (PARAMCD = 'BOR')
  , var    = CRIT1FL
  , event  = "Y"
  , by     = TRT01P
  , covars = AGE
  , class  = SEX
  , order  = TRT01PN
  , out    = resp_or
)

Output: OUT: one row per pair and statistic (BY: the group, VERSUS: the group it is compared with); GROUPS is BY VERSUS (and WITHIN).

%logrank#

The log-rank test of a time between groups (PROC LIFETEST, STRATA), one record per subject: pval. With strata= the stratified log-rank test (STRATA strata / GROUP=by): each combination of the values of the strata variables one stratum. With one group there is no test: the value is missing.

Procedure: PROC LIFETEST (STRATA)
Proven: the known answer of the method logrank (src/methods.ts, worked out by tools/method_answers.R; node tools/methods_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/logrank.sas (12NOV2026)

%logrank(
    data   = adtte
  , where  = (PARAMCD = 'TTDE')
  , var    = AVAL
  , by     = TRT01A
  , strata = STRAT1 STRAT2
  , out    = ttde_test
)

Output: OUT: one row, of all groups together (GROUPS is blank).

%lsmeans#

A linear model of a variable, one record per subject (PROC MIXED without a random effect: least squares): covariates, class factors and the groups. The differences of the LS means of pairs of groups (diff, diff_se, lcl, ucl, pval) and, when stats= asks for them, the LS mean of each group (lsmean, lsmean_se).

Procedure: PROC MIXED (LSMEANS)
Proven: the known answer of the method lsmeans (src/methods.ts, worked out by tools/method_answers.R; node tools/methods_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/lsmeans.sas (12NOV2026)

%lsmeans(
    data   = adqsadas
  , where  = (PARAMCD = 'ACTOT' and AVISIT = 'Week 24')
  , var    = CHG
  , by     = TRT01P
  , covars = BASE
  , class  = SITEGR1
  , pairs  = all
  , order  = TRT01PN
  , out    = pairs
)

Output: OUT: the rows of the differences (BY: the group, VERSUS: the group it is compared with) and of the LS means (VERSUS blank); GROUPS is BY VERSUS WITHIN.

%meanplot#

The mean (or the median) of a variable at each visit in each group, with its error bars (SE, SD, the confidence interval of the mean, or Q1 and Q3 of the median), joined by lines; the groups a little apart on the x axis so that their bars do not overlap; with ntable=yes the n of each visit and group under the axis. One panel per value of panel=.

Procedure: PROC MEANS, PROC SGPANEL
Proven: its known answer in tools/figures_sas.ts (means, SE, SD, limits worked out by hand); not yet run on a SAS server
File: sas/macros/meanplot.sas (13NOV2026)

%meanplot(
    data    = advs
  , columns = t14_7_1_cols
  , where   = (ANL01FL = 'Y' and AVISITN > 0)
  , y       = CHG
  , x       = AVISITN
  , by      = TRTA
  , panel   = PARAM
  , ntable  = yes
  , xformat = VISIT.
  , ref     = 0
  , file    = &spl_out./figures/f14_7_1
  , out     = f14_7_1
)

Output: OUT: one row per panel, group and visit: PANEL and BY, KIND (mean or median), X (the visit), Y (the statistic), LOWER and UPPER (the ends of the error bar; missing with one value for se, sd, ci), N. OUTPAR: one row per drawing parameter in the order of the table: PARAMETER ($32) its name, VALUE ($500) its value as drawn (a label in quotes). FILE: <file>.png and <file>.rtf. The quantiles are those of PROC MEANS QNTLDEF=5 (R type 2). Draw again: plot= (an out= of %meanplot, changed or not) is drawn as it is, without a statistic (error bars where LOWER has a value), and par= (an outpar= of %meanplot) gives the drawing parameters: columns, by, panel, connect, ntable, xformat, xlabel, ylabel, ref, file, width, height. They change the picture, not the numbers; the others compute, and with plot= a NOTE says each one given is not used. A drawing parameter given in the call wins over its row of par=, the row over the default of the table.

%merge_subject#

Copy variables from another data set by subject (USUBJID). The other data set has one row per subject; the rows of DATA stay as they are, in their order.

Procedure: PROC SORT, MERGE
Proven: the fixtures of the step kinds copy, period (src/adam_steps.ts; node tools/adam_steps_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/merge_subject.sas (13NOV2026)

%merge_subject(
    data = advs
  , from = adam.adsl
  , vars = AGE SEX TRTP=TRT01P TRTPN=TRT01PN
)

Output: DATA: the variables of FROM of its subject on every row, in the order of its rows; a subject not in FROM gets missing values.

%mmrm#

The mixed model for repeated measures, one record per subject and visit (PROC MIXED, REPEATED): at one visit the differences of the LS means of pairs of groups (diff, diff_se, lcl, ucl, pval) and, when stats= asks for them, the LS mean of each group (lsmean, lsmean_se). When the covariance structure does not converge the next one of fallback= is tried.

Procedure: PROC MIXED (REPEATED, LSMEANS)
Proven: the known answer of the method mmrm (src/methods.ts, worked out by tools/method_answers.R; node tools/methods_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/mmrm.sas (13NOV2026)

%mmrm(
    data     = adqsadas
  , where    = (PARAMCD = 'ACTOT' and ANL01FL = 'Y')
  , var      = CHG
  , by       = TRT01P
  , visit    = AVISIT
  , at       = "Week 24"
  , class    = SITEGR1 AVISIT
  , model    = BASE SITEGR1 TRT01P AVISIT TRT01P*AVISIT BASE*AVISIT
  , type     = un
  , fallback = toep
  , ddfm     = bw
  , pairs    = all
  , order    = TRT01PN
  , out      = mmrm
)

Output: OUT: as %lsmeans: the rows of the differences (BY, VERSUS) and of the LS means.

%param_wide#

The values of several parameters side by side: one record per group of BY that has a record of every parameter of PARAMS, with the variables VARS of each parameter as <PARAMCD>_<VAR> (WEIGHT_AVAL). Two records of one parameter in a group are an ERROR.

Procedure: PROC SORT, MERGE
Proven: the fixtures of the step kind computed (src/adam_steps.ts; node tools/adam_steps_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/param_wide.sas (12NOV2026)

%param_wide(
    data   = work.advs
  , params = WEIGHT HEIGHT
  , vars   = AVAL
  , by     = USUBJID AVISITN
  , out    = work._bmi
)

Output: OUT: one record per group of BY that has a record of every parameter: BY, then <PARAMCD>_<VAR> for each parameter and variable. Two records of one parameter in a group are an ERROR that names the group.

%pick#

Give each subject the value of its first (or last) record in another data set.

Procedure: PROC SORT, MERGE
Proven: the fixtures of the step kinds derive.from.one, derive.from.extreme, ice (src/adam_steps.ts; node tools/adam_steps_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/pick.sas (13NOV2026)

%pick(
    data   = adsl
  , from   = ex
  , value  = EXSTDT
  , order  = EXSTDT EXSEQ
  , which  = first
  , format = date9.
  , new    = TRTSDT
)

Output: DATA: NEW on every row of a subject: the value of its first (or last) record in FROM; missing for a subject without a record. With unique=yes two records of a subject are an ERROR that names the subject.

%pool#

Pool small groups: a group (e.g. a site) that has fewer than MIN subjects in one of the groups of BY (e.g. a treatment) goes into one pooled group.

Procedure: PROC FREQ, PROC SUMMARY, MERGE
Proven: the fixtures of the step kind pool (src/adam_steps.ts; node tools/adam_steps_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/pool.sas (12NOV2026)

%pool(
    data   = adsl
  , var    = SITEID
  , by     = TRT01P
  , min    = 3
  , length = $3
  , new    = SITEGR1
  , into   = 900
)

Output: DATA: NEW: the value of VAR, or INTO for a group of VAR with fewer than MIN subjects in one of the groups of BY. The rows keep their order.

%print#

Print a table or a listing: one PROC REPORT into an RTF file and a monospace listing beside it. A table is printed from its display data set (%layout), a listing from its records. The BY columns come first, a repeated value shown once; the VAR columns follow, numbers right, text left as written. ODS wraps and pages the RTF; the rows of one TOGETHER value stay on one page of the listing when they fit in ROWS. The parameters of a table can come from the setup of %layout (setup=): %print(&name.) prints its display data set again after a change; a parameter that is not in the setup is added after it: %print(&name., file=...). A keyword given twice stops the call (SAS): change it in %layout.

Procedure: PROC REPORT
Proven: its cases of node tools/print_sas.ts (PRINT SAS PASS: 18 cases) and the tables and listings of the dummy study SYNHTN01 (flat programs 12 of 12); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/print.sas (12NOV2026)

%print(
    data     = t14_2_1_display
  , by       = row
  , var      = c:
  , together = grp
  , label    = %STR(row=" ")
  , file     = &spl_out./tlf/t14_2_1
)

Output: FILE: <file>.rtf (style journal, landscape) and <file>.lst (the LISTING destination), under the titles and footnotes of the program; no row: the text of nodata=.

%profileplot#

The profile of each subject: one thin line per subject through its values by visit, in the colour and the line pattern of its group; with mean=yes the mean of each group at each visit as a thick line with markers. One panel per value of panel=.

Procedure: PROC MEANS, PROC SGPANEL
Proven: its known answer in tools/figures_sas.ts (the values and the means worked out by hand); not yet run on a SAS server
File: sas/macros/profileplot.sas (13NOV2026)

%profileplot(
    data    = adlbc
  , columns = t14_6_cols
  , where   = (PARAMCD = 'ALT' and ANL01FL = 'Y')
  , x       = ADY
  , y       = AVAL
  , by      = TRTA
  , mean    = no
  , ref     = 40
  , file    = &spl_out./figures/f14_6_1
  , out     = f14_6_1
)

Output: OUT: one row per point: PANEL, BY and SUBJECT, KIND (subject: a value of a subject; mean: the mean of a group, SUBJECT blank), X, Y, N (the values of a mean). OUTPAR: one row per drawing parameter in the order of the table: PARAMETER ($32) its name, VALUE ($500) its value as drawn (a label in quotes). FILE: <file>.png and <file>.rtf.

%proportion#

The subjects in each category of a variable in each group, their percent of the subjects of the group, and the confidence interval of that percent by a named method (PROC FREQ, BINOMIAL): n, pct, lcl, ucl (the limits in percent, as pct).

Procedure: PROC FREQ (BINOMIAL)
Proven: the known answer of the method propci (src/methods.ts, worked out by tools/method_answers.R; node tools/methods_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/proportion.sas (12NOV2026)

%proportion(
    data   = adrs
  , denom  = adsl
  , where  = (PARAMCD = 'BOR')
  , var    = AVALC
  , by     = TRT01A
  , levels = AVALC 'CR' 'PR'
  , ci     = clopper-pearson
  , out    = bor_ci
)

Output: OUT: one row per group, category and statistic; GROUPS is BY VAR.

%rate#

The exposure-adjusted incidence rate of each group: the subjects with a record (n; PROC FREQ), the sum of the exposure of the subjects of the group (exposure; PROC MEANS), the rate n / exposure times per= (rate) and its exact Poisson limits (lcl, ucl: the chi-square quantiles of 2n and 2(n + 1) degrees of freedom, halved, divided by the exposure; CINV).

Procedure: PROC FREQ, PROC MEANS, CINV
Proven: the known answer of the methods eair, eventrate (src/methods.ts, worked out by tools/method_answers.R; node tools/methods_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/rate.sas (13NOV2026)

%rate(
    data     = adae
  , denom    = adsl
  , where    = (TRTEMFL = 'Y')
  , by       = TRT01A
  , exposure = TRTDUR / 365.25
  , out      = ae_rate
)

Output: OUT: one row per group and statistic; GROUPS is BY.

%rateratio#

A count model of a count with its exposure, one record per subject (PROC GENMOD, the logarithm of the exposure the offset): Poisson or negative binomial (the dispersion k by maximum likelihood), on the groups, the covariates and the class factors. The rate ratio of pairs of groups from one fit (irr), its Wald confidence limits (lcl, ucl) and the Wald p-value of the pair (pval).

Procedure: PROC GENMOD (LSMEANS)
Proven: the known answer of the methods poisson, negbin (src/methods.ts, worked out by tools/method_answers.R; node tools/methods_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/rateratio.sas (12NOV2026)

%rateratio(
    data     = adexac
  , var      = AVAL
  , by       = TRT01P
  , exposure = TRTDUR / 365.25
  , covars   = BASE
  , dist     = negbin
  , order    = TRT01PN
  , out      = exac_rr
)

Output: OUT: one row per pair and statistic (BY: the group, VERSUS: the group it is compared with); GROUPS is BY VERSUS (and WITHIN).

%relchange#

New records of the change from the record before, relative to the normal range (the pilot SAP 8.3: a change larger than half the normal range is abnormal). For each record: a record of the parameter PREFIX + PARAMCD whose AVAL is (value - the value of the record before) / ((upper - lower limit) / 2), rounded as ROUND; ANRIND H above +LIMIT, L below -LIMIT, else N.

Procedure: PROC SORT, a DATA step (LAG)
Proven: the fixtures of the step kind relchange (src/adam_steps.ts; node tools/adam_steps_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/relchange.sas (12NOV2026)

%relchange(
    data  = work.adlbc
  , value = LBSTRESN
  , out   = work._relchange
)

Output: OUT: has the variables of DATA. AVAL has a value on a scheduled visit (AVISITN not missing) when the value, the value before and both limits are there and the upper limit is above the lower one; it is missing on the first record of a group. PARAMCD keeps its length in DATA: it must hold the new codes (the generated programs give it the length of the specification, which lists them).

%risk#

The comparison of two proportions of each pair of groups, one record per subject (PROC FREQ): the 2 x 2 table of the two groups (the group first) and the event (event=) against the rest. measure=rd: the risk difference (rd) with the interval of ci=; measure=rr: the relative risk (rr), measure=or: the odds ratio (or), each with the Wald interval of its logarithm; the p-value of Pearson's chi-square test without continuity correction (pval). With strata= the Mantel-Haenszel risk difference (COMMONRISKDIFF, CL=MH) and the CMH test.

Procedure: PROC FREQ (RISKDIFF, RELRISK, CHISQ; COMMONRISKDIFF, CMH)
Proven: the known answer of the methods riskdiff, riskratio, oddsratio, perterm (src/methods.ts, worked out by tools/method_answers.R; node tools/methods_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/risk.sas (13NOV2026)

%risk(
    data  = adrs
  , where = (PARAMCD = 'BOR')
  , var   = CRIT1FL
  , event = "Y"
  , by    = TRT01P
  , order = TRT01PN
  , ci    = newcombe
  , out   = resp_rd
)

Output: OUT: one row per pair and statistic (BY: the group, VERSUS: the group it is compared with, and WITHIN); GROUPS is BY VERSUS WITHIN.

%scatterplot#

The values of two variables against each other (a shift plot of baseline and post baseline values, two measures of a subject), one colour and marker per group; with fit=linear the least squares line of each group (its slope, intercept and R-square in out=), with fit=loess the loess curve of SGPLOT (its smoothing chosen by SGPLOT, not in out=); identity=yes the line y = x. One panel per value of panel=.

Procedure: PROC REG, PROC SGPANEL
Proven: its known answer in tools/figures_sas.ts (the points, and the slope, intercept and R-square worked out by hand); not yet run on a SAS server
File: sas/macros/scatterplot.sas (13NOV2026)

%scatterplot(
    data     = adlbc
  , columns  = t14_6_cols
  , where    = (PARAMCD = 'ALT' and AVISITN = 24)
  , x        = BASE
  , y        = AVAL
  , by       = TRTA
  , identity = yes
  , file     = &spl_out./figures/f14_6_2
  , out      = f14_6_2
)

Output: OUT: PANEL and BY, KIND, X, Y; KIND point: one row per value; with fit=linear KIND fit: one row per group with SLOPE, INTERCEPT, RSQUARE and N (X and Y missing). OUTPAR: one row per drawing parameter in the order of the table: PARAMETER ($32) its name, VALUE ($500) its value as drawn (a label in quotes). FILE: <file>.png and <file>.rtf.

%stack#

Put results one under another: the blocks of a table, in the order given.

Procedure: a DATA step (SET INDSNAME=)
Proven: the tables of the dummy study SYNHTN01 (sapling sas run: SAS RUN PASS; sapling sas check: SAS CHECK PASS, 570 of 570 numbers of R the same in SAS); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/stack.sas (12NOV2026)

%stack(
    age sex race
  , out = t14_2_01
)

Output: OUT: all rows, with RESULT (the name of the result each row comes from) and ORD (its place in the list: 1, 2, ...). %layout lays the blocks out in this order. Each variable has the longest length it has in the results: no text is cut.

%study#

Start a program: the ADaM library (read only), the output folder and the name of the program (in the footnote of each output), the options of the outputs.

Procedure: LIBNAME, OPTIONS
Proven: every program of the dummy study SYNHTN01 (sapling sas run: SAS RUN PASS; sapling sas check: SAS CHECK PASS, 570 of 570 numbers of R the same in SAS); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/study.sas (12NOV2026)

%study(
    CDISCPILOT01
  , program = t14_2_01
)

Output: CODE: the population and the records in DATA steps, the statistics macros (%describe, %freq, %ancova, ...), then %stack and %print (a figure: its figure macro).

%summarize#

Give each subject a summary of its records in another data set: the sum or the mean of a variable. A subject without a record gets a missing value; a sum of only missing values is missing.

Procedure: PROC SUMMARY, MERGE
Proven: the fixtures of the step kind derive.from.summary (src/adam_steps.ts; node tools/adam_steps_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/summarize.sas (12NOV2026)

%summarize(
    data  = adsl
  , from  = ex
  , value = DOSE_DAYS
  , stat  = sum
  , new   = CUMDOSE
)

Output: DATA: NEW on every row of a subject: the sum, mean, maximum or minimum of VALUE over its records in FROM; missing for a subject without a record.

%summarize_groups#

One record per group of BY with a summary of each variable of VARS over the records of the group: the sum, the mean, the largest or the smallest value (a missing value of a BY variable is a group of its own). A group without records has no record.

Procedure: PROC SUMMARY
Proven: the fixtures of the step kind parameter (src/adam_steps.ts; node tools/adam_steps_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/summarize_groups.sas (12NOV2026)

%summarize_groups(
    data  = _recs
  , by    = USUBJID AVISIT
  , vars  = _s1 _s2
  , stats = max sum
  , out   = _sums
)

Output: OUT: one record per group of BY with the summary of each variable of VARS (under its own name).

%table#

A table in one call: %columns, %ard (only with file=: the numbers file belongs to an output), %layout and %print, in that order, with the same parameters as the separate calls; it holds no logic of its own. It leaves the columns <data>_cols and the display data set OUT; with setup= also the global macro variable that %print(&name.) prints from, for a second print after a change of OUT. Without file= it prints nothing: change OUT, then %print(&name., file=...).

Procedure: none of its own (those of %columns, %ard, %layout, %print)
Proven: its cases of node tools/print_sas.ts (the same .lst and the same numbers file as the separate calls); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/table.sas (13NOV2026)

%table(
    data  = t14_2_1
  , n     = adsl
  , var   = trt01a
  , order = trt01an
  , total = yes
  , file  = &spl_out./tlf/t14_2_1
  , setup = t14_2_1_print
)

Output: OUT: the display data set of %layout; DATA: <data>_cols, the columns of %columns; FILE: <file>.rtf and <file>.lst of %print and the numbers file of %ard (with file=); SETUP: with setup=, the global macro variable of that name with the parameters of %print for OUT.

%table_ae#

A table of adverse events (or other occurrences) in one call: the subjects with a record in each level of var= (system organ class, preferred term), with their percent of the population. The macro writes the long program of the table (the population, the records, one %freq, %stack, %table) and runs it; it computes nothing itself. The result <name>_ae, the table <name>, <name>_cols and <name>_display stay, and the setup <name>_print. show= gives the long program.

Procedure: none of its own (those of %freq, %table)
Proven: node tools/tables_sas.ts (the same display data set as the long program, on the made-up study SYNHTN01); ran on SAS 9.4 M9, 2026-11-13
File: sas/macros/table_ae.sas (13NOV2026)

%table_ae(
    data  = adam.adae
  , adsl  = adam.adsl
  , pop   = (saffl = 'Y')
  , where = (trtemfl = 'Y')
  , var   = aebodsys aedecod
  , by    = trt01a
  , order = trt01an
  , sort  = frequency
  , label = "Treatment-emergent adverse events"
  , file  = &spl_out./tlf/t14_5_2
)

Output: DATA: <name>_ae (the result), <name> (the table), <name>_cols, <name>_display, <name>_pop and <name>_rec; FILE: <file>.rtf, <file>.lst and the numbers file (with file=); SETUP: the global macro variable <name>_print; SHOW: the long program.

%table_model#

A table of a continuous endpoint in one call: the descriptive statistics of the variables of describe= at the visit of at=, then the LS means and the differences of pairs of groups from an ANCOVA (%lsmeans, method=ancova) or from a mixed model for repeated measures over all visits (%mmrm, method=mmrm). The macro writes the long program of the table and runs it; it computes nothing itself. The results <name>_<variable> and <name>_ancova or <name>_mmrm, the table <name>, <name>_cols and <name>_display stay, and the setup <name>_print. show= gives the long program.

Procedure: none of its own (those of %describe, %lsmeans, %mmrm, %table)
Proven: node tools/tables_sas.ts (the same display data set as the long program, on the made-up study SYNHTN01); ran on SAS 9.4 M9, 2026-11-13
File: sas/macros/table_model.sas (13NOV2026)

%table_model(
    data     = advs
  , where    = (paramcd = 'SYSBP' AND anl01fl = 'Y' AND avisitn > 0)
  , covars   = base
  , at       = "Week 12"
  , describe = base aval chg
  , by       = trtp
  , order    = trtpn
  , pairs    = "Drug X 10 mg" "Placebo"
  , file     = &spl_out./tlf/t14_3_1
)

Output: DATA: <name>_<variable> (the descriptive results), <name>_ancova or <name>_mmrm, <name> (the table), <name>_cols, <name>_display, <name>_at with at=, and <name>_pop and <name>_rec with adsl= or where=; FILE: <file>.rtf, <file>.lst and the numbers file (with file=); SETUP: the global macro variable <name>_print; SHOW: the long program.

%table_response#

A table of a binary response in one call: the subjects with the event and their percent with its interval in each group (%proportion), the comparison of pairs of groups (%risk: risk difference, relative risk or odds ratio, with its test), and with covars= the odds ratio adjusted for them (%logistic). The macro writes the long program of the table and runs it; it computes nothing itself. The results <name>prop, <name><measure> and <name>_adj, the table <name>, <name>_cols and <name>_display stay, and the setup <name>_print. show= gives the long program.

Procedure: none of its own (those of %proportion, %risk, %logistic, %table)
Proven: node tools/tables_sas.ts (the same display data set as the long program, on the made-up study SYNHTN01); ran on SAS 9.4 M9, 2026-11-13
File: sas/macros/table_response.sas (13NOV2026)

%table_response(
    data   = adrs
  , var    = avalc
  , by     = trtp
  , order  = trtpn
  , pairs  = "Drug X 10 mg" "Placebo"
  , diffci = newcombe
  , covars = base
  , file   = &spl_out./tlf/t14_3_2
)

Output: DATA: <name>prop, <name><measure> and <name>_adj (the results), <name> (the table), <name>_cols, <name>_display, and <name>_pop and <name>_rec with adsl= or where=; FILE: <file>.rtf, <file>.lst and the numbers file (with file=); SETUP: the global macro variable <name>_print; SHOW: the long program.

%table_shift#

A shift table in one call: the subjects in each pair of categories, the category at baseline against the category after it (the two variables of var=), within the levels of within= (parameter, visit). The macro writes the long program of the table (the population, the records, one %freq, %stack, %table) and runs it; it computes nothing itself. The result <name>_shift, the table <name>, <name>_cols and <name>_display stay, and the setup <name>_print. show= gives the long program.

Procedure: none of its own (those of %freq, %table)
Proven: node tools/tables_sas.ts (the same display data set as the long program, on the made-up study SYNHTN01); ran on SAS 9.4 M9, 2026-11-13
File: sas/macros/table_shift.sas (13NOV2026)

%table_shift(
    data   = adlbc
  , where  = (anl01fl = 'Y' AND avisitn > 0)
  , within = paramcd avisit
  , levels = 'LOW' 'NORMAL' 'HIGH'
  , by     = trt01a
  , order  = trt01an
  , file   = &spl_out./tlf/t14_8_1
)

Output: DATA: <name>_shift (the result), <name> (the table), <name>_cols, <name>_display, and <name>_pop and <name>_rec with adsl= or where=; FILE: <file>.rtf, <file>.lst and the numbers file (with file=); SETUP: the global macro variable <name>_print; SHOW: the long program.

%table_summary#

A summary table (demographics, baseline) in one call: you name the variables, with # after a categorical one. The macro writes the long program of the table (the population, one %describe per continuous variable, one %freq per categorical variable, %stack, %table) and runs it; it computes nothing itself. The results <name>_<variable>_summ, the table <name>, the columns <name>_cols and the display data set <name>_display stay, and the setup <name>_print, for a change and a second %print. show= gives the long program, to read it or to go on from it by hand.

Procedure: none of its own (those of %describe, %freq, %table)
Proven: its cases of node tools/print_sas.ts (the same .lst and the same numbers file as the long program); ran on SAS 9.4 M9, 2026-11-13
File: sas/macros/table_summary.sas (13NOV2026)

%table_summary(
    data  = adsl
  , where = (saffl = 'Y')
  , vars  = age agegr1#agegr1n sex# race# ethnic#
  , by    = trt01a
  , order = trt01an
  , total = yes
  , file  = &spl_out./tlf/t14_2_1
)

Output: DATA: <name>_<variable>_summ (one result per variable), <name> (the table), <name>_cols, <name>_display, and <name>_pop with where=; FILE: <file>.rtf, <file>.lst and the numbers file (with file=); SETUP: the global macro variable <name>_print for %print(&<name>_print.); SHOW: the long program.

%table_tte#

A time-to-event table in one call: the Kaplan-Meier summary of each group (%km), the log-rank test (%logrank) and the hazard ratio of pairs of groups from a Cox model (%cox). The macro writes the long program of the table and runs it; it computes nothing itself. The results <name>_km, <name>_lr and <name>_cox, the table <name>, <name>_cols and <name>_display stay, and the setup <name>_print. show= gives the long program.

Procedure: none of its own (those of %km, %logrank, %cox, %table)
Proven: node tools/tables_sas.ts (the same display data set as the long program, on the made-up study SYNHTN01); ran on SAS 9.4 M9, 2026-11-13
File: sas/macros/table_tte.sas (13NOV2026)

%table_tte(
    data  = adam.adtte
  , adsl  = adam.adsl
  , pop   = (saffl = 'Y')
  , where = (paramcd = 'TTAE')
  , by    = trt01a
  , order = trt01an
  , pairs = "Drug X 10 mg" "Placebo"
  , file  = &spl_out./tlf/t14_6_1
)

Output: DATA: <name>_km, <name>_lr and <name>_cox (the results), <name> (the table), <name>_cols, <name>_display, <name>_pop and <name>_rec; FILE: <file>.rtf, <file>.lst and the numbers file (with file=); SETUP: the global macro variable <name>_print; SHOW: the long program.

%table_visit#

A table by visit in one call (vital signs, laboratory values): the descriptive statistics of each variable of vars= in each group, parameter and visit. The macro writes the long program of the table (the population, the records, one %describe per variable, %stack, %table) and runs it; it computes nothing itself. The results <name>_<variable>, the table <name>, <name>_cols and <name>_display stay, and the setup <name>_print. show= gives the long program.

Procedure: none of its own (those of %describe, %table)
Proven: node tools/tables_sas.ts (the same display data set as the long program, on the made-up study SYNHTN01); ran on SAS 9.4 M9, 2026-11-13
File: sas/macros/table_visit.sas (13NOV2026)

%table_visit(
    data  = adam.advs
  , adsl  = adam.adsl
  , pop   = (saffl = 'Y')
  , where = (anl01fl = 'Y')
  , vars  = aval chg
  , by    = trt01a
  , order = trt01an
  , keep  = param paramn avisitn
  , post  = chg
  , file  = &spl_out./tlf/t14_7_1
)

Output: DATA: <name>_<variable> (one result per variable), <name> (the table), <name>_cols, <name>_display, <name>_pop and <name>_rec; FILE: <file>.rtf, <file>.lst and the numbers file (with file=); SETUP: the global macro variable <name>_print; SHOW: the long program.

%termflags#

The flag of each row of occurrence data for every subject: one record per subject of DENOM (with its group BY) and per term (each combination of TERMS in the records), FLAG "Y" when the subject has a record of the term and "N" when not. With threshold= only the terms whose percent of the subjects of a group reaches it in at least one group. It computes no statistic: %risk then compares the groups on FLAG within each term (within=TERMS), one record per subject, the subjects of the group the denominator.

Procedure: PROC SORT, MERGE, PROC SUMMARY
Proven: the known answer of the method perterm (src/methods.ts, worked out by tools/method_answers.R; node tools/methods_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/termflags.sas (12NOV2026)

%termflags(
    data      = adae
  , denom     = adsl
  , where     = (TRTEMFL = 'Y')
  , terms     = AEBODSYS AEDECOD
  , by        = TRT01A
  , threshold = 5
  , out       = ae_flags
)

Output: OUT: one record per subject and term.

%tte#

One time-to-event parameter in one call: for each subject the first event, else the latest censoring. With end= an at-risk period from START to END: only the events and censorings in it count, else the subject is censored at the end of the period.

Procedure: PROC SORT, MERGE
Proven: the fixtures of the step kinds tte and tte.period (src/adam_steps.ts; node tools/adam_steps_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/tte.sas (12NOV2026)

%tte(
    subjects      = work.adsl
  , events        = adam.adlb
  , event         = PARAMCD in ('ALT', 'AST') and R2A1HI >= 3
  , date          = ADT
  , priority      = PARAMCD
  , start         = TRTSDT
  , end           = DCUTDT
  , censor_date   = DCUTDT
  , seq           = ASEQ
  , censor_desc   = "Cutoff date"
  , censor_source = "ADSL"
  , event_desc    = "ALT/AST >= 3x ULN"
  , source        = "ADLB"
  , paramcd       = "DIL3X"
  , out           = work.dil3x
)

Output: OUT: one record per subject with an event, a censoring or (with end=) a period: the variables of SUBJECTS, PARAMCD, ADT, CNSR, AVAL (ADT - START + 1), ATM (with time=) and the values above, in the order of SUBJECTS. With end= a subject without START or without END gets no record and a WARNING that names it (USUBJID=); a subject whose START is after its END gets no record. Two records of a subject in SUBJECTS are an ERROR.

%ttest#

The two-sample t-test of each pair of groups, one record per subject (PROC TTEST, BY each pair, the group first): the difference of the means (diff: the group minus the group it is compared with), its standard error (diff_se), its confidence limits (lcl, ucl) and the two-sided p-value (pval): Student's with the pooled variance (variances=equal) or Welch's with Satterthwaite's df (variances=unequal).

Procedure: PROC TTEST
Proven: the known answer of the method ttest (src/methods.ts, worked out by tools/method_answers.R; node tools/methods_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/ttest.sas (13NOV2026)

%ttest(
    data  = advs
  , where = (PARAMCD = 'SYSBP' and AVISIT = 'Week 12')
  , var   = CHG
  , by    = TRT01P
  , order = TRT01PN
  , out   = sbp_t
)

Output: OUT: one row per pair and statistic (BY: the group, VERSUS: the group it is compared with); GROUPS is BY VERSUS.

%visits#

Give each record its analysis visit by its visit: AVISIT and AVISITN. A record of a visit that is not in the list keeps them missing.

Procedure: DATA steps (SET POINT=)
Proven: the fixtures of the step kind windows.visit (src/adam_steps.ts; node tools/adam_steps_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/visits.sas (13NOV2026)

%visits(
    data   = adlb
  , visits = "SCREENING" 0 "Baseline"
           # "WEEK 2"    2 "Week 2"
)

Output: DATA: AVISIT and AVISITN on each record of a visit in the list; missing on the others.

%wilcoxon#

The Wilcoxon rank-sum test of each pair of groups, one record per subject (PROC NPAR1WAY, BY each pair, the group first): the two-sided p-value of the normal approximation with the continuity correction (pval; CORRECT=), the Hodges-Lehmann estimate of the shift (estimate: the group minus the group it is compared with) and its asymptotic (Moses) confidence limits (lcl, ucl; HL).

Procedure: PROC NPAR1WAY (WILCOXON, HL)
Proven: the known answer of the method wilcoxon (src/methods.ts, worked out by tools/method_answers.R; node tools/methods_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/wilcoxon.sas (13NOV2026)

%wilcoxon(
    data  = adqs
  , where = (PARAMCD = 'PAIN' and AVISIT = 'Week 12')
  , var   = CHG
  , by    = TRT01P
  , order = TRT01PN
  , out   = pain_w
)

Output: OUT: one row per pair and statistic (BY: the group, VERSUS: the group it is compared with); GROUPS is BY VERSUS.

%windows#

Put each record in its analysis window by its study day: AVISIT, AVISITN, AWLO, AWHI, AWTARGET, AWRANGE, AWU (DAYS) and AWTDIFF (the distance to the target day). A record in no window keeps them missing. The first window (in the order of AVISITN) that holds the day is taken.

Procedure: DATA steps (SET POINT=)
Proven: the fixtures of the step kind windows.day (src/adam_steps.ts; node tools/adam_steps_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/windows.sas (13NOV2026)

%windows(
    data    = advs
  , windows = 0 "Baseline" .  1  1  "<=1"
            # 4 "Week 4"   2  42 28 "2-42"
            # 8 "Week 8"   43 70 56 "43-70"
  , day     = ADY
)

Output: DATA: AVISIT, AVISITN, AWLO, AWHI, AWTARGET, AWRANGE, AWU and AWTDIFF on each record in a window; missing on a record in no window.

%worst#

The records of each subject at its worst level (the maximum severity or toxicity grade): for each subject and each value of BY (a term, or nothing: all its records), only the records whose VAR is the highest of LEVELS, in their order from the least to the most severe. A record whose VAR is not one of LEVELS (a missing grade) is never the worst and is left out. It computes no statistic: %freq then counts each subject once in each row, at its worst level.

Procedure: PROC SORT, a DATA step (BY, FIRST.)
Proven: the known answer of the method catsumm with worst= (src/methods.ts, counted by hand; R ars_worst gives the same, test/safety-r.test.ts; node tools/methods_sas.ts); ran on SAS 9.4 M9, 2026-11-12
File: sas/macros/worst.sas (12NOV2026)

%worst(
    data   = adae
  , where  = (TRTEMFL = 'Y')
  , var    = AESEV
  , by     = AEBODSYS AEDECOD
  , levels = 'MILD' 'MODERATE' 'SEVERE'
  , out    = ae_worst
)

Output: OUT: the records of DATA at the worst level of each subject and row.