tabtools

Version 1.0.0 | 2026-10-20

[

Quick Start · Installation · Commands · Worked Examples · Stata-Tools · Report a bug

tabtools turns descriptive, model, survival, rate, and composite results into publication-ready Excel and GitHub-Flavored Markdown tables. Its commands share one set of output options, so the same rendered table can go to a workbook, a Markdown report, a CSV file, or a Stata frame.

Quick Start

After installing, build a baseline table from the synthetic cohort in the repository's _data/ folder and write it as Markdown (the replace treated line is the demo's: it makes treatment depend on age, sex, diabetes, and hypertension, so the two arms differ at baseline):

use "https://raw.githubusercontent.com/tpcopeland/Stata-Tools/main/_data/cohort.dta", clear
merge 1:1 id using "https://raw.githubusercontent.com/tpcopeland/Stata-Tools/main/_data/treatment.dta", nogen keep(match)
merge 1:1 id using "https://raw.githubusercontent.com/tpcopeland/Stata-Tools/main/_data/comorbidities.dta", nogen keep(master match) nolabel
foreach v in diabetes hypertension {
    replace `v' = 0 if missing(`v')
}
set seed 20261005
replace treated = runiform() < invlogit(-2.4 + 0.05 * (index_age - 58) - 0.3 * female + 2.2 * diabetes + 2.0 * hypertension)
table1_tc, by(treated) vars(index_age contn %5.1f \ female bin \ diabetes bin \ hypertension bin) title("Table 1. Baseline Characteristics") markdown("table1.md")

table1.md holds exactly this (its ### heading shown in bold here):

Table 1. Baseline Characteristics

No. (Column %) or Mean±SD	SSRI (N=8,929)	SNRI (N=6,071)	p-value
Age at cohort entry (years)	55.9±13.4	62.0±12.4	<0.001
Female sex	5,530 (62)	3,442 (57)	<0.001
Diabetes	2,566 (29)	4,359 (72)	<0.001
Hypertension	2,765 (31)	4,282 (71)	<0.001

_{The first table of demo/demo_markdown_report.md, written by demo/demo_tabtools.do, which then appends a crosstab and a puttab table to the same file with mdappend. The demo also merges the outcome file and builds variables this table does not use; they do not change it. The same call prints the table in the Results window, and xlsx() or frame() would send it to a workbook or a frame as well.}

Features

• One set of output options. A table command can write any of the targets it supports, xlsx(), csv(), markdown(), frame(), and, for model and effect tables, a graph-ready eplotframe(), from the same rendered table in one call.
• Tables from the models you already ran. regtab and effecttab format the active collect result of regress, logistic, margins, teffects and similar commands, and comptab combines their frames into one composite table.
• Small-cell disclosure control. smallcells(#) in table1_tc, desctab, and crosstab applies primary, complementary, and dependent-result suppression before any output is written, so every sink receives the same redacted table.
• Confidence levels that are checked. An explicit level() that conflicts with the level recorded in a collection or a saved strate file is an error, not a relabelled interval.
• Shared house style. tabtools set and tabtools get hold session defaults for font, font size, borders, colours, digits, and p-value emphasis, and permanent saves them to a runnable profile.

Requirements

• Stata 17 or later
• No mandatory user-written Stata packages
• regtab and effecttab need an active Stata collect result created by commands such as regress, logistic, margins, or teffects; they format and arrange that collection rather than fitting a model
• survtab needs data declared with stset; stratetab reads .dta files written by Stata's strate, output() command
• Optional: the eplot package from Stata-Tools, for forest-plot output from comptab and hrcomptab and for plotting the eplotframe() outputs of regtab and effecttab (the table commands create those graph-ready frames without it)

Installation

capture ado uninstall tabtools
net install tabtools, from("https://raw.githubusercontent.com/tpcopeland/Stata-Tools/main/tabtools") replace

If another copy is earlier on the adopath, inspect ado dir and remove only the tabtools package before reinstalling. The public distribution includes the command files, help files, package metadata, and table-of-contents entry; it does not include the repository's demo fixtures or generated demo workbooks.

Commands

The suite contains 16 public commands. The version column is the minimum Stata release for that command.

Command	Stata	Purpose
table1_tc	17+	Baseline Table 1 by group, with continuous, categorical, missingness, tests, and standardized mean differences
desctab	17+	Direct access to the consolidated descriptive engine used by table1_tc
crosstab	17+	Two-way categorical tables with percentages, tests, and 2x2 effect measures
regtab	17+	Render active regression collections with estimates, confidence intervals, statistics, and optional plot frames
effecttab	17+	Render active margins or teffects results, or a supplied effect matrix
tabcell	17+	Format one publication cell — estimate (CI), p-value, n (%), e/n (%), or median (Q1, Q3) — from a fit, lincom/nlcom, a matrix row, or numbers, or fill a string column
outtab	17+	Binary outcomes by exposure: events/N (%) per group plus one ratio column per model specification
survtab	17+	Kaplan–Meier survival, event, risk-set, median, RMST, and group-difference tables
stratetab	17+	Convert saved strate, output() rate files into rate and rate-ratio tables
ratetab	17+	Events, person-time, and rates by group from stset or event/exposure data, with exact, Poisson, or cluster-robust intervals
hrcomptab	17+	Compatibility wrapper for comptab rate-scaffold mode
comptab	17+	Combine model frames vertically or interlock them with a rate scaffold
puttab	17+	Put variables, a frame, or a matrix into a formatted workbook or Markdown table
stacktab	17+	Stack or place blocks from an existing workbook into a new worksheet
tabtools	17+	Inspect and set shared fonts, digits, borders, colors, and persistent profiles; session destinations (tabtools set workbook/markdown); fit-time counts (tabtools fitcount)
tabtools_tips	17+	Open or print a compact recipe reference for the suite

How It Works

Most table commands follow the same three-stage pattern: calculate or receive results, render a table in Results, then optionally export or expose that same rendered structure. The export targets are independent, so a command can write Excel, CSV, Markdown, a Stata frame, or an eplot-ready frame in one call when that command supports the target.

xlsx() is the main Excel option and excel() is retained as a synonym on commands that support both names. sheet() selects the worksheet. open opens an Excel target after writing it and therefore requires an xlsx() or using target. csv() exports the visible table data for commands that offer it; markdown() writes GitHub-Flavored Markdown, and mdappend appends to an existing Markdown file rather than replacing it.

frame(name[, replace]) stores the rendered table in a Stata frame for later composition or inspection. eplotframe(name[, replace]) stores graph-ready estimates, confidence limits, p-values, labels, and model identifiers for the model/effect commands that support it. comptab consumes compatible model/effect frames and, with rateframe(), a rate frame plus model frames; hrcomptab is the compatibility wrapper for that mode.

For plot-ready composition, create each display frame and its numeric companion in the same command call. Their pair identities must agree. Recreate older saved pairs with the current regtab or effecttab before requesting a composite plot frame.

Shared formatting options include font(), fontsize(), borderstyle(), headershade, headercolor(), zebra, zebracolor(), title(), footnote(), boldp(), and highlight() where supported. A fresh session resolves the shared baseline as Arial 10-point text with thin borders, while command-specific precision and display defaults are listed below.

Workbook cell types

Every exported workbook cell is written as text, including cells that look numeric. This is deliberate: the majority of published table cells are composite or annotated strings — 5,351 (60), 0.82 (0.69, 0.98), <0.001, 0.54***, Reference — and the commands render each cell to its final string before any sink runs. Re-deriving a numeric type by reparsing the rendered string would have to guess, and would silently convert an annotated cell into a number that no longer matches the table.

The practical consequence is that Excel will not sum, chart, or numerically sort a column of an exported sheet without a conversion step in Excel. For a numeric payload, use frame() (Stata frames, with the underlying numeric columns where a command provides them), the r() matrices such as r(table) and r(cutoff_table), or eplotframe() for plotting.

Choosing a Workflow

Need	Start with	Add or follow with
Baseline characteristics by one group variable	table1_tc	smd, test, xlsx(), frame()
Direct access to the Table 1 engine	desctab	the same syntax and options as table1_tc
Two categorical variables	crosstab	rowpct, colpct, or, rr, rd, fisher, trend, or cochran
Regression model collection	regtab	stats(), keep()/drop(), eplotframe(), or frame()
Marginal effects or treatment effects	effecttab	from(), type(), clean, full, or eplotframe()
Survival probabilities or RMST	survtab	times(), by(), median, riskset, rmst(), or difference
Incidence rates from strate	stratetab	outlabels(), rateratio, or a later hrcomptab
Several model/effect results	comptab	compatible source frames and rows()
Rates plus model estimates	comptab	rateframe(), rows(), and optional forest
A dataset, frame, or matrix as a table	puttab	using, frame(), matrix(), varlabels, or noheader
Existing Excel blocks	stacktab	blocks(), layout(hstack), append, or sheetreplace

Worked Examples

1. Baseline table

sysuse auto, clear
table1_tc price mpg weight rep78, by(foreign) smd test frame(table1, replace) xlsx("table1.xlsx")

table1_tc detects the supplied variables when vars() is omitted. Use vars() for explicit row types such as contn, conts, cat, bin, and their extended forms; smd requires by(), and clear is available when the rendered table should replace the data in memory.

To protect exact counts below a publication threshold in table1_tc, desctab, or crosstab, add smallcells(#). Primary cells are shown as <#, complementary cells as ≥#, and dependent statistics as Suppressed. A missing or negative count that is not printed but follows from printed cells and sample sizes is protected in the same way. Every requested sink receives the same already-redacted table. After safety is certified, the engine deterministically removes complementary markers one at a time whenever the protected counts remain non-exact. Each remaining ≥# is therefore individually necessary in the final table, although the result is not guaranteed to use the globally smallest possible number of complementary markers.

table1_tc rep78, by(foreign) vars(rep78 cat) total(after) smallcells(5) frame(table1_safe, replace)

2. Shared formatting profile

Use the controller to inspect the suite, set explicit formatting defaults, and save them for reuse in another session:

tabtools, detail category(models)
tabtools set font Arial
tabtools set fontsize 10
tabtools set borderstyle thin, permanent profile("tabtools_project.do")
tabtools use using "tabtools_project.do"
tabtools get

Use tabtools, list for the compact catalog, or replace category(models) with another catalog category. The profile contains ordinary tabtools set commands and can be version controlled with a project.

3. Consolidated descriptive engine

sysuse auto, clear
desctab price mpg weight rep78, by(foreign) smd test frame(descriptive, replace) xlsx("descriptive.xlsx")

desctab is the consolidated implementation behind table1_tc. It accepts the same variable-type, inference, suppression, weighting, formatting, and output options; table1_tc is the stable user-facing frontend.

4. Regression and effects

sysuse auto, clear
collect clear
collect: logistic foreign mpg weight
regtab, frame(regression, replace) eplotframe(regression_plot, replace) xlsx("regression.xlsx")

collect clear
collect: margins, at(mpg=(15 25 35))
effecttab, frame(effects, replace) xlsx("effects.xlsx")

regtab and effecttab format active collections; they do not fit models. Both can also feed a composite table, and their eplotframe() output can be plotted by the optional eplot package.

5. Categorical tables

sysuse auto, clear
crosstab foreign rep78, rowpct fisher smallcells(5) xlsx("crosstab.xlsx")

crosstab defaults to column percentages and uses a sparse-cell exact test when appropriate. With smallcells(#), it protects the count block and margins, withholds percentages whose numerator or denominator is protected, and suppresses count-dependent tests and effect measures.

6. Survival and rates

webuse drugtr, clear
stset studytime, failure(died)
survtab, times(1 2 3) by(drug) median riskset xlsx("survival.xlsx")

webuse diet, clear
stset dox, failure(fail)
strate hienergy, per(1) output(rate_hienergy, replace)
stratetab, using("rate_hienergy") outcomes(1) xlsx("rates.xlsx")

survtab reports Kaplan–Meier quantities at the requested times. stratetab expects the saved strate files in exposure-major, outcome-minor order and uses outcomes() to interpret that file sequence.

7. Composite model table

sysuse auto, clear
collect clear
collect: regress price mpg weight
regtab, frame(model, replace)
collect clear
collect: regress price mpg weight length
regtab, frame(model2, replace)
comptab model model2, rows("1 2 \ 1 2") xlsx("composite.xlsx")

The source frames must contain compatible row labels. rows() takes one numeric row specification per source frame, separated by \; here 1 2 \ 1 2 selects the first two rows from both frames. Use relabel() or section() when the displayed row names or section structure need to be changed; use hrcomptab when one of the source frames is a stratetab rate frame.

8. Direct table output

sysuse auto, clear
puttab price mpg weight using "selected.xlsx", sheet("Table") varlabels
puttab price mpg weight, markdown("selected.md")

puttab accepts a current-data varlist, frame(name), or matrix(name) as its one source. An Excel using target is required for workbook output, while Markdown-only output does not need a workbook.

Demo

Run demo/demo_tabtools.do from a repository checkout to regenerate the workbooks and the Markdown report below; it uses the repository's _data/ fixtures and writes under demo/. These files are repository documentation assets and are not part of the net install payload.

File	What it holds
demo/demo_markdown_report.md	Three tables appended to one Markdown file with mdappend: table1_tc, crosstab with an odds ratio, and puttab records
demo/demo_table1.xlsx	16 sheets: table1_tc with totals, weights, wtcompare, statistics and format variants, missingness, explicit fonts, small cells, and cellreplace()
demo/demo_desctab.xlsx	10 sheets: descriptive tables (events, mean and SD, median and IQR) and small cells
demo/demo_regtab.xlsx	18 sheets: logistic, Cox, mixed, Poisson, and GEE tables, multi-model layouts, keep()/drop(), addrow(), cformat(), a flat frame with reftop, transpose, cellnote(), and tabtools fitcount counts
demo/demo_regtab_models.xlsx	10 sheets: model families (multinomial and ordered logit, OLS, probit, negative binomial, GLM Poisson, panel random effects, quantile, zero-inflated, hurdle)
demo/demo_effecttab.xlsx	5 sheets: teffects ATE, a comparison, margins and AMEs, and cformat()
demo/demo_comptab.xlsx	7 sheets: two source tables and compact, renamed, formatted, and flat composites
demo/demo_hrcomptab.xlsx	1 sheet: a rate-and-hazard-ratio composite
demo/demo_stratetab.xlsx	2 sheets: rates from strate output, plain and with cformat()
demo/demo_ratetab.xlsx	5 sheets: exact, Poisson, and region-clustered rates, formatted rates, and crude and adjusted hazard ratios beside the rates through comptab, rateframe() allmodels keyed
demo/demo_outtab.xlsx	2 sheets: events/N by treatment with crude and adjusted risk ratios for four outcomes, and a two-panel odds-ratio version
demo/demo_survtab.xlsx	3 sheets: Kaplan–Meier survival, RMST, and cumulative incidence
demo/demo_crosstab.xlsx	8 sheets: cross-tabulations with measures, styling, trend tests, row percentages, and small cells
demo/demo_puttab.xlsx	6 sheets: matrix, frame, and collapsed sources, stacked tables with hlines()/boldrows(), panel(), and spanheader()
demo/demo_stacktab.xlsx	5 sheets: two source blocks, a stacked composite, a side-by-side layout, and stacktab frames()
demo/demo_tabcell.xlsx	1 sheet: a column of tabcell cells written through a tabtools set workbook session destination

The table1_tc, desctab, and crosstab workbooks each contain paired smallcells(5) examples. The Small Cells Primary sheets use a 2×2 table with counts 2/3/3/2, so only below-threshold cells are hidden as <5. The Small Cells Complement sheets use counts 2/8/6/4, which also hide reconstructive cells as ≥5. The Small Cells Primary Mode sheets rerun the 2/8/6/4 table with smallcells(5, primary), which hides only the two printed counts below 5.

The optional forest-plot demo is demo/demo_tabtools_eplot.do. It regenerates the two graphs below from regtab and comptab workflows and needs the optional eplot and tc_schemes packages in the checkout.

Output	Command focus
	regtab, eplotframe() plotted with eplot
	comptab, forest

Command Reference

The command help files are the authoritative reference for abbreviations and complete option parsing. The syntax and defaults below summarize the public contracts.

table1_tc

table1_tc [varlist] [if] [in] [fweight], [by(varname) vars(string) format(string) percformat(string) nformat(string) iqrmiddle(string) sdleft(string) sdright(string) gsdleft(string) gsdright(string) percent missing pdp(#) highpdp(#) test statistic excel(string) xlsx(string) sheet(string) title(string) clear percent_n percsign(string) spacelowpercent extraspace slashN total(string) catrowperc varlabplus headerperc font(string) fontsize(#) borderstyle(string) wt(varname) smd smdtype(string) smdpair(g1 g2[, values|labels]) footnote(string) open boldp(#) zebra highlight(#) headershade frame(string) smdthreshold(#) nosmdhighlight headercolor(string) zebracolor(string) csv(string) markdown(string) mdappend missingsummary smallcells(#[, primary]) nosmallcells cellreplace(string) dots wtcompare wtn nopvalue]

table1_tc is Stata 17+ and accepts frequency weights. Without vars(), it infers row types from the varlist; the default display formats are %2.0f, %5.0f, and %12.0fc for common continuous, percentage, and count cells, with pdp(3), highpdp(2), and an SMD threshold of 0.1. The Excel sheet defaults to Table 1; nosmdhighlight (or smdthreshold(-1)) disables SMD highlighting, and clear replaces the current dataset with the table. smallcells(#) requires an integer threshold of at least 3 and protects exact disclosure within one invocation; it does not certify anonymization or account for linkage across separate releases. smallcells(#, primary) masks only printed counts 1 to #−1 as <# (with their percentages) and adds no complementary suppression; the footnote names the mode. cellreplace("row label" column "text" [\ ...]) overwrites body cells and errors unless each names exactly one cell. puttab, varlabels consumes the embedded header row of a frame() table, so no drop in 1 is needed. With three or more by() groups, smd compares the first two groups by default (smdtype(pair)) and says so in every output: the column header is SMD (A vs B) in the console, Excel, CSV, Markdown, frame(), and clear, and a note naming the comparison joins footnote(), r(smdnote), and the frame characteristic _dta[tabtools_smdnote]. smdpair(g1 g2) picks the two groups by by() value or label; a number that is one group's value and another group's label is refused unless , values or , labels says which is meant. smdtype(population) reports the largest absolute difference between a group mean and the overall mean in overall-sample SDs (McCaffrey et al. 2013; header Pop. SB), and smdtype(maxpair) the largest absolute pairwise SMD on one shared scale, the root-mean of the group variances (header Max SMD). A group with no values for a variable leaves that variable's statistic blank; it is never computed over fewer groups. Two-group output is unchanged.

desctab

desctab [varlist] [if] [in] [fweight], [the same options as table1_tc]

desctab is the direct engine entry point and has the same syntax, defaults, outputs, and stored results as table1_tc. Existing desctab collect-formatter syntax from tabtools 1.x is intentionally unsupported.

crosstab

crosstab rowvar colvar [if] [in] [fweight=exp], [xlsx(string) excel(string) colpct rowpct totalpct or rr rd trend cochran exact fisher label missing level(#) digits(#) title(string) footnote(string) font(string) fontsize(#) borderstyle(string) headershade headercolor(string) zebracolor(string) boldp(#) zebra csv(string) markdown(string) mdappend frame(string) smallcells(#[, primary]) nosmallcells open]

crosstab is Stata 17+, accepts numeric categorical variables and frequency weights, and defaults to column percentages, the current c(level), and session digits or 1. smallcells(#) requires an integer of at least 3 and protects counts, released margins, dependent percentages, tests, and requested measures before any sink runs. or, rr, and rd require a 2x2 table; trend and cochran are separate ordered-trend tests; exact and fisher are synonyms. Numeric level order, not value-label order, determines the requested 2x2 measures.

regtab

regtab, [xlsx(string) excel(string) sheet(string) sep(string) models(string) coef(string) nointercept keepintercept noreffects stats(string) relabel(string) digits(#) footnote(string) open zebra headershade highlight(#) boldp(#) cdisc font(string) fontsize(#) borderstyle(string) stars starslevels(numlist) headercolor(string) zebracolor(string) csv(string) markdown(string) mdappend frame(string) eplotframe(name[, replace]) keep(string) drop(string) labelmatch dimnonsig factorlabel refcat(string) omitlabel(string) emptylabel(string) notestlabel(string) absentlabel(string) cnslabel(string) cutlabels(string) addrow(string) compact nopvalue pdp(#) highpdp(#) labelwidth(#) level(#) cformat(%fmt) reftop cellnote(string) mincount(#) transpose exposurelabel(string) cilabel(string) plabel(string) addcol(string) statlabels(string) collabels(string)]

regtab is Stata 17+ and renders the active collect result. The sheet defaults to Regression, digits to the session setting or 2, sep() to , , pdp(3), highpdp(2), refcat() to Reference, omitlabel() to Omitted, emptylabel() to Empty, labelwidth() to 45, and starslevels() to 0.05 0.01 0.001. Ratio-scale models receive their conventional coefficient labels and suppress intercepts automatically where appropriate; keep() and drop() are mutually exclusive. stats() accepts n, obs, aic, bic, qic, icc, ll, groups, r2, events, people, and exposure; obs prints e(N) even for survival models, and events/people/exposure work for any model after tabtools fitcount (st models fall back to e(N_fail) for events).

keep() and drop() match exact, case-sensitive raw variable or factor-component names: age cannot select stage or Age, and 2.arm cannot select 20.arm. Use labelmatch explicitly for case-insensitive display-label substring matching, for example keep(Fuel) labelmatch.

The command does not fit models and can alter the active collection's layout and styles. Explicit level() must agree with collection metadata. When a Stata version omits that metadata and level() is not supplied, regtab warns and uses the current c(level) for interval labels; supply level() if the models were fit at a different level. nopvalue hides p-value columns but does not remove p-values used by stars or highlighting.

cformat() takes a full numeric display format for the estimate and both limits (for example %12.0fc for thousands separators) and cannot be combined with digits(); a comma-decimal format is refused when the CI separator contains a comma. sep() reaches every output. frame(name, flat) writes one row per body line with printed headers as variable labels, ready for puttab, varlabels. reftop prints a non-lowest reference level first in its block. cellnote("row label" # "text" [\ ...]) replaces one model × term cell and errors unless the label matches exactly one row. mincount(#), after tabtools fitcount, terms, blanks a term whose level has fewer than # events, no events, or no estimable variance, printing emptylabel() (default – under mincount()). transpose prints models as rows and terms as columns for specification tables.

notestlabel() labels a factor level the model holds but could not estimate (an empty cell, or a level estimated with zero or missing variance); it defaults to the emptylabel() text. When collect records every constrained cell as empty (nbreg, zinb, intreg, some streg fits), regtab decides the reference level from the active fit's own e(b) (b. is the reference, o. is omitlabel()), or otherwise from the model's command line with fvexpand, so ib#., ib(last)., and fvset base resolve as at fit time; a base that only a later fvset names is not called the reference. Under mincount(), a level of a factor the model includes but that no observation of the model's sample holds, such as a level an if condition leaves out, prints absentlabel() (default blank), while a level the sample holds but the model could not estimate prints notestlabel(); both are counted in r(N_absent). Base, omitted, empty, and constrained cells are read from the fit's own notes, which tabtools fitcount stores, so a variable dropped after the fit keeps its reference label, and a base that cannot be decided prints a note rather than a guessed reference. A coefficient fixed by a constraint prints its value with cnslabel() (default (constrained)) in place of the interval. stintreg fits get the TR or HR header their metric implies. A column whose fit failed (collected with no estimates and no counts) is accepted under mincount() and left blank with a note, and a model collected twice from the same e() is flagged. frame(name, flat) labels each column with the statistic header it prints under (HR, 95% CI), keeping the model name in char c#[tabtools_block] and the full header in char c#[tabtools_header]; puttab, blockheader prints the model names back as a spanning row. frame(name, flat keys) adds _order, _term (the raw coefficient name, such as 9.edss_t, or stat:<item> for a stats() row), _rowtype, and a _state# per model column (est, ref, omit, notest, absent, empty, constrained, masked), so a single cell can be replaced by coefficient name, a row inserted before a term, or a rule drawn above a statistics row with ordinary Stata commands before puttab; puttab skips the key variables unless they are named. A row given as addrow("label" values, after(term)) is placed inside the body right after the coefficient row term (raw names as keep() takes them) or after the last level of factor term. stats() items take a display format and small-cell masking, e(N, %9.0fc) and e(N_fail, mincell(5)) (masked cells counted in r(N_stats_masked)), combine two results as one cell, e(N_fail|N_sub) printing 31 (48), and text("label" value1 value2 ...) adds a row of text per model. Every occurrence of the same e() result takes the strictest mincell() given for it, and maskwith(names) masks rows that would give a masked count back (for example people without an event next to events) wherever any row of the group is masked, counted in r(N_stats_linked). Under transpose, collabels(name "label" ...) relabels the term columns (eq:name for one equation's column) and addcol("label" values, after(term)) places a column after a term's columns. Labels and values in addrow() and text(), and the text of statlabels(), exposurelabel(), coef(), cilabel(), plabel(), models(), refcat(), omitlabel(), notestlabel(), cutlabels(), collabels(), and cellnote(), absentlabel(), cnslabel(), addcol(), and sep(), are written as typed: a $name is not expanded.

effecttab

effecttab, [xlsx(string) excel(string) sheet(string) sep(string) type(string) effect(string) models(string) title(string) clean tlabels(string) footnote(string) open zebra headershade highlight(#) boldp(#) font(string) fontsize(#) borderstyle(string) full digits(#) headercolor(string) zebracolor(string) csv(string) markdown(string) mdappend frame(string) eplotframe(name[, replace]) from(name) addrow(string) pdp(#) highpdp(#) labelwidth(#) level(#) refcat(string) cformat(%fmt)]

effecttab is Stata 17+ and accepts an active margins/teffects collection or a matrix through from(). The sheet defaults to Effects, digits to 2, sep() to , , pdp(3), highpdp(2), refcat() to Reference, and labelwidth() to 45; type() and effect() are inferred when omitted. Matrix input uses 95% intervals unless level() is supplied, and collection-level provenance rules match regtab. cformat() and frame(name, flat) behave as in regtab.

survtab

survtab, times(numlist) [by(varname) rmst(#) median riskset timeunit(string) reverse difference events level(#) digits(#) xlsx(string) excel(string) sheet(string) title(string) footnote(string) font(string) fontsize(#) borderstyle(string) headershade headercolor(string) boldp(#) zebra zebracolor(string) highlight(#) pdp(#) highpdp(#) csv(string) markdown(string) mdappend frame(string) open addrow(string)]

survtab is Stata 17+ and requires stset data. Frequency weights declared by stset use replication semantics consistently for Kaplan-Meier estimates, event and risk-set counts, and Greenwood RMST variance. Probability and importance weights are rejected before output. The default time unit is years, the sheet is Survival, the confidence level is c(level), digits are the session setting or 1, and pdp()/highpdp() default to 3/2. by() adds group columns; median, riskset, events, and rmst() add corresponding quantities; difference reports group 1 minus group 2 RMST when exactly two groups are supplied. reverse reports 1 − KM and is not a competing-risks estimator.

stratetab

stratetab, using(string) outcomes(integer) [xlsx(string) excel(string) sheet(string) title(string) outlabels(string) outcomeids(string) explabels(string) digits(#) eventdigits(#) pydigits(#) unitlabel(string) ptunit(string) pyscale(#) ratescale(#) rateratio ratiodigits(#) footnote(string) open zebra font(string) fontsize(#) borderstyle(string) headershade headercolor(string) zebracolor(string) csv(string) markdown(string) mdappend frame(string) level(#) cformat(%fmt) sep(string) smallcells(#) nosmallcells masktext(string) zeroexact zerocells(dash|blank[, persontime])]

stratetab is Stata 17+ and reads .dta files produced by strate, output(). Pass the output() filename stem to using(); stratetab adds the .dta suffix automatically. outcomes() is required and must divide the number of input files; files are interpreted as all outcomes for exposure 1, then all outcomes for exposure 2, and so on. Defaults are sheet Results, digits(1), eventdigits(0), pydigits(0), unitlabel("1,000"), pyscale(1), ratescale(1000), and ratiodigits(2). Rate confidence-level metadata must be present and consistent, or be supplied explicitly with level(). smallcells(#) suppresses a rate with 1 to #−1 events together with its person-time and honours the session default; zeroexact prints the exact one-sided upper limit for zero-event cells instead of the default dash. A row with zero person-time (_Y = 0) has no rate and is left empty, its rate ratio and reference cell included, and is counted in r(N_nopt); a row with events but no person-time is refused.

hrcomptab

hrcomptab rateframe, modelframes(framelist) rows(string) [rownames(string) outcomemap(string) xlsx(string) excel(string) sheet(string) csv(string) markdown(string) mdappend frame(string) eplotframe(name[, replace]) forest eplotoptions(string) open title(string) footnote(string) effect(string) reflabel(string) font(string) fontsize(#) borderstyle(string) zebra headershade headercolor(string) zebracolor(string)]

hrcomptab is a compatibility wrapper for comptab, rateframe(). It preserves the existing syntax, defaults, outputs, and stored results.

comptab

comptab framelist, rows(string) [rownames(string) xlsx(string) excel(string) sheet(string) title(string) footnote(string) compact separator(numlist) section(string) relabel(string) font(string) fontsize(#) borderstyle(string) open zebra headershade highlight(#) boldp(#) headercolor(string) zebracolor(string) csv(string) markdown(string) mdappend frame(string) eplotframe(name[, replace]) forest eplotoptions(string) labelwidth(#) cformat(%fmt) cisep(string)]

comptab modelframes, rateframe(name) rows(string) [rownames(string) effect(string) reflabel(string) outcomemap(string) allmodels keyed modelonly cformat(%fmt) cisep(string) xlsx(string) excel(string) sheet(string) title(string) footnote(string) font(string) fontsize(#) borderstyle(string) open zebra headershade headercolor(string) zebracolor(string) csv(string) markdown(string) mdappend frame(string) eplotframe(name[, replace]) forest eplotoptions(string)]

comptab is Stata 17+ and combines compatible regtab/effecttab source frames. Supply exactly one of rows() or rownames(); the sheet defaults to Composite and labelwidth() to 45. Without rateframe(), separator(), section(), relabel(), and compact control vertical composition. With rateframe(), the model frames are interlocked with a stratetab scaffold and effect(), reflabel(), and outcomemap() become available. The two option families are mutually exclusive. hrcomptab forwards to this rate mode. In rate mode, allmodels (or an outcomemap() naming several frames per outcome) prints one effect column per model; keyed places model rows on their exact block and level identity, allows sections that carry rates only, and refuses a row it cannot place exactly; modelonly lists model rows with no rate section (spline read-outs, nonlinearity tests) after the table. cformat() re-renders estimates from the numeric companions, cisep() sets the interval separator, and frame(name, flat) writes the flat frame.

puttab

puttab [varlist] [if] [in] [using filename.xlsx], [frame(string) matrix(name) sheet(string) title(string) footnote(string) font(string) fontsize(#) borderstyle(string) headercolor(string) zebracolor(string) zebra headershade digits(#) nformat(%fmt) varlabels noheader noembedheader hlines(numlist) vlines(numlist) boldrows(numlist) csv(string) markdown(string) mdappend open noheadershade panel(varname) panelheader(varlist|strings) panelinline noindent spanheader(string) blockheader]

puttab is Stata 17+ and accepts exactly one source: a current-data varlist, frame(), or matrix(). The default sheet is Table and digits default to the session setting or 2; using is required for Excel output, while Markdown-only output can omit it. varlabels uses variable labels and noheader suppresses the header row. nformat(%12.0fc) writes every integer-valued column with thousands separators; a column whose own display format ends in fc keeps its separators without it, with digits() still setting its decimals. hlines() draws a rule above the listed data rows, vlines() a rule right of the listed columns, and boldrows() bolds the listed data rows, so a second table stacked in the same source can be ruled off with its own bold header row. panel(varname) starts a bold, ruled heading row (the value label or value) whenever the panel variable changes and indents the rows beneath it; panelheader(varlist) repeats a per-panel header row, and literal text such as panelheader("" "Events" "PY") writes the same header under every panel; noindent keeps panel row labels flush, and panelinline writes each panel heading into the first cell of its panelheader() row, so heading and column headers share one row (the first header cell must be blank). blockheader, for a regtab flat frame, adds a row of model names merged over each model's columns (read from char c#[tabtools_block]), and variables marked as regtab frame keys are skipped unless named. spanheader("label" first/last [\ ...]) adds a second header row with merged spanning labels. After tabtools set workbook or tabtools set markdown, using and markdown() may be omitted; the first write replaces the file and later writes add sheets or append.

stacktab

stacktab using outbook.xlsx, blocks(blockspec) sheet(sheetname) [layout(string) title(string) note(string) footnote(string) columnmerge style(string) borders(string) spacing(#) csv(string) markdown(string) mdappend frame(string) display append sheetreplace]

stacktab [using outbook.xlsx], frames(name ["label"] [\ ...]) sheet(sheetname) [title(string) footnote(string) csv(string) markdown(string) mdappend]

stacktab is Stata 17+ and reads an existing .xlsx workbook. blocks() identifies the source ranges, sheet() identifies the output worksheet, and layout() defaults to vertical stacking; hstack places blocks horizontally. The default title cell is A1, the first table starts at B2, and spacing() defaults to 0. append and sheetreplace control an existing target sheet and cannot be used together. frames() stacks in-memory frames such as table1_tc output, each under its own panel heading, through puttab, panel().

tabcell

tabcell est [coef] [, lincom nlcom matrix(M row) cols(string) b(#) ll(#) ul(#) se(#) eform format(%fmt) cformat(%fmt) sep(string) level(#) missing(text)]
tabcell p, p(#) [pdp(#) highpdp(#) pstyle(table|footnote|Pfootnote) missing(text)]
tabcell n, n(#) [mincell(#) nformat(%fmt) missing(text)]
tabcell np, n(#) d(#) [ci(exact) level(#) sep(string) nocount mincell(#) nformat(%fmt) pformat(%fmt) missing(text)]
tabcell enp, e(#) n(#) [mincell(#) nformat(%fmt) pformat(%fmt) missing(text)]
tabcell iqr, median(#) q1(#) q3(#) [format(%fmt) sep(string) missing(text)]
tabcell rate, e(#) pt(#) per(#) [ci(exact|poisson) level(#) format(%fmt) digits(#) sep(string) mincell(#) missing(text)]
tabcell form, options local(name) global(name)
tabcell form [if] [in], generate(newvar) options

tabcell is Stata 17+ and returns one formatted cell in r(cell), for example 1.23 (1.01 to 1.50) with sep(" to "). Its p-value text follows the same rule as regtab. A missing or non-finite value is an error unless missing() supplies the text, so a failed fit never prints . (., .). mincell(#) prints counts 1 to #−1 as <#; tabcell rate and tabcell np, nocount, which print no count, print – for a masked cell instead. With generate(), the column form fills a string variable from numeric variables in one pass; otherwise tabcell never changes the data. local(name) and global(name) also store the cell text in a macro, so a cell is made and kept in one line; text with quotes, backquotes, or $ is stored as typed, and a failed call clears the named macro. pstyle(Pfootnote) writes a p-value as P = 0.012, P < 0.001. tabcell np, ci(exact) adds the exact binomial (Clopper–Pearson) interval of the percentage, 2 (10.0; 1.2, 31.7), with limits equal to cii proportions, exact (0 when n = 0, 100 when n = d); non-integer counts are refused. nocount prints the percentage and its interval alone, 10.0 (2.1, 26.5). tabcell rate prints a rate per per() units of person-time with the same intervals as ratetab: exact Poisson limits by default (the upper limit −ln(α/2)/pt for zero events) or log-scale limits with ci(poisson); 27 events in 2,200 person-years per 1,000 print 12.3 (8.1, 17.9) with the default exact limits. digits() sets the decimals of any form that takes format().

ratetab

ratetab groupvars [if] [in] [, events(varlist) exposure(varlist) per(#) ci(exact|poisson|cluster(varname)) level(#) smallcells(#) nosmallcells masktext(string) excludemasked zerocells(dash|blank[, persontime]) cformat(%fmt) digits(#) pydigits(#) pyscale(#) ptunit(string) sep(string) outlabels(string) explabels(string) unitlabel(string) saving(filename[, replace]) stratetab_options]

ratetab is Stata 17+ and computes events, person-time, and rates by each grouping variable, from stset data or from events() and exposure(). ci(exact) (the default) gives exact Poisson limits, and a zero-event cell shows its exact upper limit. ci(poisson) gives log-scale Wald limits, and ci(cluster(id)) fits a saturated Poisson model with an exposure offset and robust variance clustered on id, for recurrent events within people; the number of clusters is reported. It renders through stratetab, so its frame() feeds comptab, rateframe() directly. zerocells(dash|blank) prints a zero-event cell's count and rate as – or empty, and the persontime suboption withholds its person-time too. A level with no person-time for an outcome has no rate: its cells are left empty in every output and it is counted in r(N_nopt). saving() writes one row per level with the grouping variables under their own names, types, and labels; a grouping variable named like a saved column (group, level, ...) is saved as g_<name> and recorded in _dta[ratetab_renamed]. stratetab's rateratio is refused: it compares independent exposure groups, and ratetab's grouping variables split one sample, so for rate ratios between the levels of one variable fit poisson _d i.var, exposure(pt) irr with pt = _t - _t0. Without ptunit(), pyscale(), or an stset scale() of 365 or 365.25, a note says that person-time is labelled in person-years by default; a nonzero rate or limit that rounds to 0 at the chosen decimals is noted too.

outtab

outtab outcomes [if] [in], exposure(varname) [models(string) modellabels(string) estimator(string) eform minevents(#) mintext(string) smallcells(#) nosmallcells panels(varlist) obsprefix(name) grouplabels(string) ratiolabel(string) format(%fmt) sep(string) nonconvtext(string) failtext(string) droptext(string) frame(string) xlsx(string) excel(string) sheet(string) title(string) footnote(string) csv(string) markdown(string) mdappend borderstyle(string) headershade font(string) fontsize(#)]

outtab is Stata 17+ and tabulates binary outcomes by a binary exposure: events/N (%) in each group and one ratio column per model in models() (each a covariate list, "" for crude), fitted with estimator() (for example poisson, irr vce(cluster id)). Adjusted models are complete-case. A fit that fails, does not converge, drops observations beyond its complete-case sample, or has too few exposed events (minevents()) prints its text instead of a ratio. The caller's e() is restored.

tabtools

tabtools [, list detail category(string)]
tabtools set key value [, permanent profile(string)]
tabtools set clear [, permanent profile(string)]
tabtools get
tabtools use [using filename] [, profile(string)]
tabtools set {workbook|markdown} filename
tabtools set headershade {on|off}
tabtools set smallcells # [primary]
tabtools set borderstyle {default|thin|medium|academic|clear}
tabtools set sessionkey clear
tabtools query
tabtools fitcount, events(varname) [people(varname) exposure(varname) terms name(collection)]

tabtools is Stata 17+. list displays the command catalog, detail adds descriptions, and category() filters descriptive, models, rates, survival, composite, export, general, or all. set keys are font, fontsize, borderstyle, headercolor, zebracolor, digits, and boldp; tabtools set fontsize accepts 6–72 points, digits accept 0–6, and border styles are default, thin, medium, and academic. permanent writes a runnable profile in the Stata PERSONAL directory, and profile() selects an alternate profile path; use loads a profile for the session. Session keys (workbook, markdown, headershade, smallcells) last for the Stata session, are echoed whenever a command uses them, and lose to any explicit option; tabtools query lists them. borderstyle is both a settings key and a session key: every command with a borderstyle() option uses it when that option is not given, tabtools query reports it, tabtools set borderstyle clear removes it, and permanent saves it to a profile. tabtools fitcount runs right after a collect: fit and stores events, people, person-time, and per-level event counts on e(sample) for regtab; it refuses a fit that does not match the collected model, frequency or importance weights, and svy, subpop().

tabtools_tips

tabtools_tips [, open]

tabtools_tips is Stata 17+ and prints a compact recipe reference; open opens its help file. It has no stored results.

Key Options

Output targets

• xlsx(filename) writes an Excel workbook; excel(filename) is a compatibility synonym where listed in command syntax.
• sheet(name) selects the Excel sheet. Defaults are Table 1 for table1_tc/desctab, Crosstab for crosstab, Regression for regtab, Effects for effecttab, Survival for survtab, Results for stratetab, Composite for comptab/hrcomptab, and Table for puttab. stacktab requires an explicit sheet name.
• csv(filename) writes the visible table data for commands that support CSV output, and must name a .csv file. Titles and footnotes are not additional CSV columns.
• The workbook, csv(), and markdown() must name different files; paths are compared after resolving them against the working directory and ignoring case, and a collision is refused before anything is written.
• markdown(filename) writes GitHub-Flavored Markdown and replaces an existing file. mdappend appends when the target exists and creates it otherwise. Cell, header, title, and footnote text is written literally: it is never macro-expanded, and \ | * _, backtick, < > & [ ] ~ are backslash-escaped so data never render as emphasis, HTML, entities, links, strikethrough, or code. Leading spaces in the row-label (first) column are written as   entities, so indented categorical and factor-level rows keep their hierarchy after GFM trims cell whitespace; value cells are trimmed.
• frame(name[, replace]) stores the rendered table; eplotframe(name[, replace]) stores graph-ready model/effect results.
• open requires an Excel target and asks Stata to open the written workbook.

Formatting

• Use font(), fontsize(), and the explicit border, header, and zebra options to control formatting.
• borderstyle() accepts default, thin, medium, or academic; a fresh session's baseline is thin.
• headershade, headercolor(), zebra, and zebracolor() control header and alternating-row appearance.
• title() and footnote() add report text; command-specific note() or section() options are documented with the commands that support them.
• boldp(#) and highlight(#) emphasize statistically notable cells where supported; nopvalue hides p-value columns in regtab without discarding p-values used for styling.

Selection and precision

• digits() controls decimal display where available; command defaults are listed in the command reference, and tabtools set digits provides the session default.
• pdp() and highpdp() control ordinary and small p-value display in commands that show p-values.
• keep() and drop() are mutually exclusive in commands that offer both; models(), coef(), relabel(), cutlabels(), and addrow() provide command-specific selection or annotation.
• level() controls confidence intervals only where the command accepts it. Collection and saved-rate commands reject conflicting or unavailable confidence-level metadata.

Suite controller and profile options

Option	Applies to	Purpose
list	tabtools display mode	Show the public command catalog as a simple list
detail	tabtools display mode	Add command descriptions to the catalog
category(string)	tabtools display mode	Filter the catalog by descriptive, models, rates, survival, composite, export, general, or all
font(string)	tabtools set font	Set the default font family
fontsize(#)	tabtools set fontsize	Set the default font size in points; valid values are 6–72
permanent	tabtools set and tabtools set clear	Save the resulting defaults to a runnable profile on disk
profile(filename)	tabtools set ..., permanent and tabtools use	Choose an alternate profile file instead of the default tabtools_profile.do in Stata's PERSONAL directory

Set borderstyle, headercolor, and zebracolor directly with their corresponding tabtools set keys.

Stored Results

All result-producing commands return output paths and dimensions when the corresponding target is requested. Names below are r() results unless noted otherwise; dynamic names use # for a model, group, outcome, or estimator index.

table1_tc

Returns r(markdown_rows), r(markdown_cols), r(Dapa), r(methods), r(varlist), r(xlsx), r(sheet), r(frame), r(markdown), and r(table). With smallcells(#), it also returns r(smallcells), r(N_primary_suppressed), r(N_secondary_suppressed), r(N_derived_suppressed), and the code matrix r(suppression); protected p-values and SMDs in r(table) are .d. With smd it returns r(smdtype) (pair, population, or maxpair) and, when a note names the comparison, r(smdnote); the r(table) column is smd for every smdtype().

desctab

Returns the same results as table1_tc.

crosstab

Returns r(N), r(ci_level), r(chi2), r(p), requested r(or), r(rr), or r(rd), trend results r(p_trend), r(chi2_trend), and r(z_trend) where applicable, plus r(markdown_rows), r(markdown_cols), r(table), r(methods), r(trend_method), r(xlsx), r(sheet), r(frame), and r(markdown). With smallcells(#), it also returns r(smallcells), suppression counts, and r(suppression); protected counts use .p/.s and dependent inferential results use .d.

regtab

Returns r(N_rows), r(N_cols), r(N_models), r(ci_level), r(markdown_rows), r(markdown_cols), r(xlsx), r(sheet), r(markdown), r(coef_label), r(methods), r(stars), r(frame), r(eplotframe), and r(table). Model-specific statistics use dynamic names such as r(n_#), r(obs_#), r(people_#), r(exposure_#), r(aic_#), r(bic_#), r(qic_#), r(icc_#), r(ll_#), and r(groups_#) where available; stats(e(name)) items return r(e_name_#). mincount() adds r(N_masked) and r(N_absent) (levels a model leaves out or could not estimate), and a stats() item with mincell() adds r(N_stats_masked), plus r(N_stats_linked) when maskwith() masks a row.

effecttab

Returns r(N_rows), r(N_cols), r(ci_level), r(markdown_rows), r(markdown_cols), r(xlsx), r(sheet), r(markdown), r(type), r(effect_label), r(methods), r(frame), r(eplotframe), and r(table).

survtab

Returns r(N_rows), r(table), r(ci_level), r(logrank_p), r(logrank_chi2), r(n_groups), r(markdown_rows), r(markdown_cols), r(by_var), r(xlsx), r(sheet), r(markdown), r(csv), r(methods), and r(frame). Group and time summaries use dynamic names such as r(median_#), r(events_#), r(atrisk_#), r(rmst_#), r(rmst_se_#), r(rmst_lb_#), r(rmst_ub_#), r(group_#_value), and r(group_#_label); two-group RMST differences use r(rmst_diff), r(rmst_diff_se), r(rmst_diff_lb), r(rmst_diff_ub), and r(rmst_diff_p).

stratetab

Returns r(N_rows), r(N_exposures), r(N_outcomes), r(ci_level), r(markdown_rows), r(markdown_cols), r(rates), r(ratios), r(xlsx), r(sheet), r(frame), r(outcome_ids), r(markdown), and r(methods), plus r(smallcells) when masking applies and r(N_nopt), the cells with zero person-time.

hrcomptab

Returns r(N_rows), r(N_outcomes), r(N_sections), r(N_modelrows), r(N_modelframes), r(ci_level), r(markdown_rows), r(markdown_cols), r(rateframe), r(modelframes), r(effect), r(xlsx), r(sheet), r(markdown), r(csv), r(frame), and r(eplotframe).

comptab

Returns r(N_rows), r(N_cols), r(N_models), r(N_frames), r(ci_level), r(markdown_rows), r(markdown_cols), r(frame), r(markdown), r(xlsx), r(sheet), r(methods), and r(eplotframe). Rate mode also returns r(N_models_per_outcome) and r(N_modelonly).

puttab

Returns r(n_rows), r(n_cols), r(n_datarows), r(source), and, when applicable, r(sheet), r(file), r(csv), r(markdown), r(markdown_rows), r(markdown_cols), r(n_panels), and r(n_spans).

stacktab

Returns r(blocks_loaded), r(rows_written), r(rows_out), r(cols_out), r(append_start), r(layout), r(sheet), r(markdown), r(book), r(table_start), r(title_cell), r(frame), r(csv), and optional r(note_row), r(markdown_rows), and r(markdown_cols). frames() returns r(n_frames) and r(frames).

tabcell

Returns r(cell); the est form adds r(estimate), r(lb), r(ub), r(level), and r(source), and r(missing) flags a cell printed from missing(). tabcell np, ci(exact) adds r(pct), r(lb), and r(ub) (percentages, missing when no interval is printed), r(level), and r(citype); tabcell rate returns r(rate), r(lb), r(ub), r(level), r(per), and r(citype). The column form returns r(N), r(N_missing), and r(varname).

ratetab

Returns the stratetab results plus r(N), r(per), r(level), r(N_zero), r(N_noci), r(N_nopt), r(ci_method), r(cluster), r(methods), the cell matrix r(estimates), and, for clustered intervals, r(clusters).

outtab

Returns r(N_rows), r(N_models), r(N_outcomes), r(N_panels), r(smallcells), r(minevents), r(estimator), r(frame), r(xlsx), the counts-and-ratios matrix r(table), and the per-fit diagnostics matrix r(fits).

tabtools

tabtools display mode returns r(commands), r(n_commands), r(version), and r(categories). set returns the changed setting, r(permanent), r(profile), r(action) when clearing, and r(key), the key removed by set key clear; get returns r(font), r(fontsize), r(borderstyle), r(headercolor), r(zebracolor), r(digits), and r(boldp). use returns r(action) = "loaded" and r(profile). query returns r(workbook), r(markdown), r(headershade), r(smallcells), r(smallcells_mode), r(borderstyle), r(workbook_fresh), and r(markdown_fresh). fitcount returns r(N), r(events), r(people), r(exposure), r(cmdset), r(collection), and, with terms, r(n_terms) and r(terms).

Assumptions and Limits

• All tabtools commands require Stata 17 or newer.
• regtab and effecttab format existing collect results. desctab and table1_tc preserve the caller's active collection while using a private collection internally.
• regtab and effecttab require confidence-level metadata to agree with an explicit level(); stratetab applies the same rule to saved strate metadata.
• table1_tc uses Stata frequency-weight syntax. Probability and importance weights are not silently treated as frequency weights; weighted tables use weighted percentages by default, and wtn requests effective counts where supported.
• Standardized mean differences require table1_tc, by(). wtcompare and wtn require wt().
• crosstab effect measures require a 2x2 table and can be undefined for zero cells. Ordered trend tests require the relevant binary or ordered variables.
• survtab requires stset; reverse is a complementary Kaplan–Meier display and does not model competing risks. RMST differences are defined for two groups.
• stratetab depends on the exact file order emitted by strate, output() and requires outcomes() to describe that order. Its rateratio matching uses exposure labels and treats the first exposure as the reference.
• comptab and hrcomptab require compatible source-frame row identifiers. Forest output is an optional eplot integration, not a required table dependency.
• ratetab, stratetab, and outtab small-cell masking covers printed counts only; it adds no complementary suppression across grouping variables or exposures.
• tabtools fitcount counts are unweighted and identify the fit by e(cmdline), e(N), and e(b); run it immediately after the collect: fit.
• stacktab reads existing .xlsx workbooks and uses Stata's Excel facilities; source blocks must identify valid worksheet ranges.
• tabtools set permanent writes a runnable profile in the user's Stata PERSONAL directory. It changes future sessions only when that profile is loaded or sourced.
• Excel output requires a writable target path, and open additionally requires a graphical Excel-capable environment. Markdown and CSV targets do not require Excel.
• smallcells(#) protects exact disclosure within one invocation of table1_tc, desctab, or crosstab. Its deterministic final pass makes the complementary set irredundant, not necessarily globally minimum. It does not certify anonymization or account for linkage across separate releases; an unsupported or unprovable layout fails before output.

References

• Optional forest plots use the eplot package when installed.

QA

QA suites and how to run them are documented in qa/README.md.

Author

Timothy P Copeland, Karolinska Institutet

Questions and bug reports: GitHub issues. Please include the first line of which tabtools and your Stata version.

To cite: Copeland TP. tabtools: Manuscript tables to Excel and Markdown. Stata-Tools, https://github.com/tpcopeland/Stata-Tools/tree/main/tabtools. Please give the version you used (which tabtools prints it).

License

MIT (see LICENSE)