tabtools

Version 1.6.4 | 2026-06-10

tabtools is a suite of Stata commands for exporting manuscript-ready tables to Excel and Markdown across descriptive summaries, regression models, treatment effects, survival analysis, diagnostic accuracy workflows, incidence rates, and composite tables. The package is organized around a shared formatting layer, so commands that come from very different analysis pipelines still produce tables that look like they belong in the same workbook or report.

Requirements

• Stata 16 or later for tabtools, tabtools_tips, table1_tc, stacktab, and simtab
• Stata 17 or later for desctab, regtab, effecttab, comptab, hrcomptab, survtab, crosstab, corrtab, diagtab, stratetab, and puttab
• desctab, regtab, and effecttab require Stata's collect framework
• survtab requires stset data, and stratetab expects saved strate, output() datasets
• eplot is optional and is required only when using comptab, forest or hrcomptab, forest

Installation

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

After installation, start with help tabtools for the suite overview and tabtools_tips for the merged quick reference and worked recipes. The older help tabtools_cheatsheet and help tabtools_cookbook topics still work as compatibility aliases.

Markdown Export

Every table-producing command accepts markdown(filename) to write the rendered table as GitHub-Flavored Markdown. Use mdappend to append several tables into one report file. Markdown can be requested alone where a workbook is not structurally required, or in the same call as xlsx()/excel(), csv(), and frame().

table1_tc age bmi sex, by(treated) markdown("table1.md")
regtab, xlsx("models.xlsx") markdown("models.md")
crosstab sex treated, markdown("tables.md")
corrtab age bmi sbp, markdown("tables.md") mdappend

Commands

Direct table builders

Post-estimation formatters

File and frame workflow builders

Styled export and assembly

puttab vs comptab vs stacktab

These three commands all produce a single combined or styled sheet, but they differ by what they read:

Workflow: puttab and stacktab form an emit-then-assemble pipeline — puttab writes each styled block to its own sheet, then stacktab combines those sheets into the final table. comptab/hrcomptab are the frame-based siblings of stacktab: reach for them when the pieces are still tabtools frames rather than exported sheets.

In short: style one raw table → puttab; combine estimation results still in frames → comptab/hrcomptab; combine sheets already in a workbook → stacktab.

Simulation studies

simtab renders and exports a publication-ready simulation performance table. For full performance analysis, Monte Carlo error theory, and diagnostic graphs (zipper, lollipop, nested-loop), use simsum or siman. simtab can read their output directly (from(simsum) / from(siman)), or compute table-grade measures itself from replication-level data — it installs and runs with neither package present. It pairs with simsum/siman (Morris, White & Crowther, Stat Med 2019), which own the numbers; simtab owns the styled table.

Suite utility

Choosing a Workflow

Persistent Defaults Profiles

tabtools set still applies defaults immediately through Stata session globals. Add permanent to save the current style to a runnable Stata profile, then load it in a later session or project with tabtools use.

tabtools set theme custom, font("Times New Roman") fontsize(10) ///
    borderstyle(academic) permanent profile("tabtools_house_style.do")
tabtools set digits 3, permanent profile("tabtools_house_style.do")

tabtools set clear
tabtools use using "tabtools_house_style.do"
tabtools get

Without profile(), tabtools set ..., permanent writes tabtools_profile.do in Stata's PERSONAL ado directory, and tabtools use reads that default profile.

Repository Checkout Demo

The rebuild demo is a repository-maintenance workflow, not part of the net install payload. It reads shared _data/ fixtures and sibling packages from a local Stata-Tools checkout, then regenerates console output, a sequential Markdown report, and 14 Excel workbooks (72 sheets total) covering every tabtools command.

From a local checkout, run:

stata-mp -b do tabtools/demo/demo_tabtools.do

Installed users should start with help tabtools and tabtools_tips; the legacy help tabtools_cheatsheet and help tabtools_cookbook topics are shipped as compatibility aliases.

Markdown report export

The demo builds demo/demo_markdown_report.md by exporting one table and appending additional tables with mdappend:

table1_tc, by(treated) ///
    vars(index_age contn %5.1f \ female bin \ diabetes bin \ hypertension bin) ///
    title("Table 1. Baseline Characteristics") ///
    markdown("tabtools/demo/demo_markdown_report.md")

crosstab treated female, or label ///
    title("Table 2. Treatment by Sex") ///
    markdown("tabtools/demo/demo_markdown_report.md") mdappend

corrtab index_age crp prior_hosp, star(0.05 0.01 0.001) ///
    title("Table 3. Correlation Matrix") ///
    markdown("tabtools/demo/demo_markdown_report.md") mdappend

puttab id index_age treated female cv_event, varlabels ///
    title("Table 4. First Six Analysis Records") ///
    markdown("tabtools/demo/demo_markdown_report.md") mdappend

Excerpt from the generated Markdown report:

### Table 1. Baseline Characteristics

| No. (Column %) or Mean (SD) | N=8,934 | N=6,066 | p-value |
| --- | --- | --- | --- |
| Age at cohort entry (years) | 58.3 (13.4) | 58.5 (13.3) | 0.24 |
| Female sex | 5,351 (59.9%) | 3,621 (59.7%) | 0.80 |
| Diabetes | 4,107 (46.0%) | 2,818 (46.5%) | 0.56 |
| Hypertension | 4,112 (46.0%) | 2,935 (48.4%) | 0.005 |

Simulation performance tables (simtab)

demo/demo_simtab.do builds a Monte Carlo study of three estimators (Unweighted, IIW, IIW + log(test)) across three scenarios and two estimands, then renders it with simtab. The Unweighted estimator is biased — its coverage is flagged off-nominal (*) and its failed fits surface through nsim() as non-convergence (Non-conv.). Run it from a local checkout:

stata-mp -b do tabtools/demo/demo_simtab.do

Compute mode summarizes the raw replications into a styled, scenario-grouped table (console preview, with the off-nominal-coverage flag and per-cell non-convergence count):

. simtab estid, estimate(est) se(se) true(truev) by(scen) sim(sim) coverage(covered) ///
      nsim(400) metrics(mean bias empse meanse coverage n nonconv) digits(3) ///
      xlsx("demo_simtab.xlsx") sheet("Scenarios") display

Simulation results by scenario (400 replications)
  +----------------------------------------------------------------------------------------------+
  | Scenario         Estimator    Mean     Bias   Emp. SE   Mean SE   Coverage     N   Non-conv. |
  |        A        Unweighted   0.142   +0.042     0.040     0.042       86%*   372          28 |
  |                        IIW   0.093   -0.007     0.039     0.042        95%   400           0 |
  |            IIW + log(test)   0.109   +0.009     0.039     0.042        96%   400           0 |
  |        B        Unweighted   0.151   +0.051     0.042     0.042       75%*   379          21 |
  |                        IIW   0.102   +0.002     0.042     0.042        96%   400           0 |
  |            IIW + log(test)   0.120   +0.020     0.039     0.042        94%   400           0 |
  |        C        Unweighted   0.157   +0.057     0.041     0.042       74%*   379          21 |
  |                        IIW   0.110   +0.010     0.040     0.042        96%   400           0 |
  |            IIW + log(test)   0.128   +0.028     0.040     0.042       92%*   400           0 |
  +----------------------------------------------------------------------------------------------+

Coverage is empirical 95% CI coverage; * flags off-nominal coverage.

The numeric plotframe() companion stores one row per cell with the raw measures and their Monte Carlo SEs — the structured source for figures, replacing the "parse a text log" boundary:

  +-----------------------------------------------------------------------------------------+
  | by_label   estimator_label    mean     bias   empse   coverage   mcse_c~e   nfail     n |
  |        A        Unweighted   0.142    0.042   0.040      0.858      0.018      28   372 |
  |        A               IIW   0.093   -0.007   0.039      0.952      0.011       0   400 |
  |        A   IIW + log(test)   0.109    0.009   0.039      0.962      0.009       0   400 |
  |       ...                                                                               |
  +-----------------------------------------------------------------------------------------+

With two estimands, Excel gets merged column-group headers (one block per estimand) and Markdown/CSV get flattened Estimand: metric columns. The demo writes the Multi-estimand sheet of demo/demo_simtab.xlsx and this Markdown report (demo/demo_simtab_report.md):

### Simulation results by scenario and estimand

| Scenario | Estimator | Marginal slope: Mean | Marginal slope: Bias | Marginal slope: Coverage | Marginal slope: N | Treatment contrast: Mean | Treatment contrast: Bias | Treatment contrast: Coverage | Treatment contrast: N |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| A | Unweighted | 0.142 | +0.042 | 86%* | 372 | 0.540 | +0.040 | 85%* | 372 |
|  | IIW | 0.093 | -0.007 | 95% | 400 | 0.493 | -0.007 | 95% | 400 |
|  | IIW + log(test) | 0.109 | +0.009 | 96% | 400 | 0.512 | +0.012 | 94% | 400 |

Ingest mode renders an already-computed per-cell summary without recomputation — from(summary) is dependency-free, and from(simsum) / from(siman) read those packages' output directly (simtab cross-validates to exact agreement with both). simtab itself installs and runs with neither package present.

Suite overview

. tabtools

──────────────────────────────────────────────────────────────────────
tabtools - Publication-Ready Table Export Suite
──────────────────────────────────────────────────────────────────────

**Descriptive Statistics**
  table1_tc    - Table 1 with automatic statistical tests
  desctab      - Format descriptive table collects
  crosstab     - Cross-tabulation with association measures
  corrtab      - Correlation matrix with significance

**Model Results**
  regtab       - Regression results from any estimation command
  effecttab    - Treatment-effect style tables from supported results

**Incidence Rates**
  stratetab    - Incidence rates from strate output

**Survival Analysis**
  survtab      - Kaplan-Meier estimates, medians, and RMST

**Diagnostic Accuracy**
  diagtab      - Sensitivity, specificity, PPV, NPV, ROC

**Composite**
  comptab      - Combine regtab/effecttab frames into one table
  hrcomptab    - Attach regtab frames to a stratetab scaffold

**Styled Export**
  puttab       - Style an in-memory dataset, frame, or matrix as one sheet
  stacktab     - Assemble multi-sheet composite Excel tables from blocks

**Simulation Studies**
  simtab       - Monte Carlo performance table

**General Purpose**
  tabtools     - Suite controller and persistent defaults
  tabtools_tips - Quick reference and worked recipes

──────────────────────────────────────────────────────────────────────
Total commands: 16

table1_tc — Baseline characteristics

. table1_tc, by(treated) ///
>     vars(index_age contn %5.1f \ female bin \ ///
>          education cat \ income_quintile cat \ ///
>          born_abroad bin \ civil_status cat \ ///
>          diabetes bin \ hypertension bin \ anxiety bin \ prior_cvd bin)

  ┌────────────────────────────────────────────────────────────────────────┐
  │                                SSRI            SNRI            p-value │
  ├────────────────────────────────────────────────────────────────────────┤
  │ No. (Column %) or Mean (SD)    N=8,934         N=6,066                 │
  ├────────────────────────────────────────────────────────────────────────┤
  │ Age at cohort entry (years)    58.3 (13.4)     58.5 (13.3)      0.24   │
  ├────────────────────────────────────────────────────────────────────────┤
  │ Female sex                     5,351 (59.9%)   3,621 (59.7%)    0.80   │
  ├────────────────────────────────────────────────────────────────────────┤
  │ Education level                                                 0.11   │
  │    Primary                     2,333 (26.1%)   1,527 (25.2%)           │
  │    Secondary                   3,530 (39.5%)   2,354 (38.8%)           │
  │    Tertiary                    3,071 (34.4%)   2,185 (36.0%)           │
  ├────────────────────────────────────────────────────────────────────────┤
  │ Disposable income quintile                                      0.73   │
  │    1                           1,778 (19.9%)   1,175 (19.4%)           │
  │    2                           1,783 (20.0%)   1,249 (20.6%)           │
  │    3                           1,769 (19.8%)   1,228 (20.2%)           │
  │    4                           1,786 (20.0%)   1,209 (19.9%)           │
  │    5                           1,818 (20.3%)   1,205 (19.9%)           │
  ├────────────────────────────────────────────────────────────────────────┤
  │ Born outside Sweden            1,362 (15.2%)   897 (14.8%)      0.44   │
  ├────────────────────────────────────────────────────────────────────────┤
  │ Marital status                                                  0.34   │
  │    Single                      2,764 (30.9%)   1,804 (29.7%)           │
  │    Married                     3,074 (34.4%)   2,162 (35.6%)           │
  │    Divorced                    1,763 (19.7%)   1,188 (19.6%)           │
  │    Widowed                     1,333 (14.9%)   912 (15.0%)             │
  ├────────────────────────────────────────────────────────────────────────┤
  │ Diabetes                       4,107 (46.0%)   2,818 (46.5%)    0.56   │
  ├────────────────────────────────────────────────────────────────────────┤
  │ Hypertension                   4,112 (46.0%)   2,935 (48.4%)    0.005  │
  ├────────────────────────────────────────────────────────────────────────┤
  │ Anxiety disorder               6,079 (68.0%)   4,148 (68.4%)    0.66   │
  ├────────────────────────────────────────────────────────────────────────┤
  │ Prior cardiovascular disease   5,002 (56.0%)   3,390 (55.9%)    0.90   │
  └────────────────────────────────────────────────────────────────────────┘

survtab — Kaplan-Meier survival summary

. survtab, times(365 730 1095 1460) by(treated) ///
>     rmst(1460) difference median timeunit(days)

                                          SSRI (N=8934)               SNRI (N=6066)               Difference
────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
 Median survival, d                       5699.0                      5660.0                      39.0
   (95% CI)                               (5617.0, 5760.0)            (5574.0, 5729.0)
 Survival probability
   365 days                               99.9%                       99.8%                       0.1 (-0.1, 0.2)
   730 days                               99.5%                       99.5%                       -0.1 (-0.3, 0.2)
   1095 days                              99.0%                       99.1%                       -0.1 (-0.4, 0.2)
   1460 days                              97.3%                       97.5%                       -0.2 (-0.8, 0.3)
 RMST (1460-d), d (95% CI)                1448.87 (1447.02, 1450.72)  1449.75 (1447.55, 1451.96)  -0.88
 Log-rank test: chi2(1) = 2.80, p = 0.094

regtab — Regression table

. collect clear
. collect: logistic treated index_age female i.education ///
>     diabetes hypertension anxiety prior_cvd
. regtab, coef("OR") noint display

                              Model
────────────────────────────────────────────────────────────────
                              OR         95% CI        p-value
 Age at cohort entry (years)  1.00       (1.00, 1.00)  0.27
 Female sex                   0.99       (0.93, 1.06)  0.84
 Education level
   Primary                    Reference
   Secondary                  1.02       (0.94, 1.11)  0.66
   Tertiary                   1.09       (1.00, 1.18)  0.053
 Diabetes                     1.01       (0.94, 1.08)  0.78
 Hypertension                 1.10       (1.03, 1.17)  0.005
 Anxiety disorder             1.00       (0.93, 1.08)  0.96
 Prior cardiovascular disease 0.98       (0.92, 1.05)  0.63

Compact layout keeps p-values but combines the point estimate and confidence interval:

. regtab, coef("OR") noint compact display

                              Model
──────────────────────────────────────────────────────────
                              OR 95% CI          p-value
 Age at cohort entry (years)  1.00 (1.00, 1.00)  0.27
 Female sex                   0.99 (0.93, 1.06)  0.84
 Education level
   Primary                    Reference
   Secondary                  1.02 (0.94, 1.11)  0.66
   Tertiary                   1.09 (1.00, 1.18)  0.053
 Diabetes                     1.01 (0.94, 1.08)  0.78
 Hypertension                 1.10 (1.03, 1.17)  0.005
 Anxiety disorder             1.00 (0.93, 1.08)  0.96
 Prior cardiovascular disease 0.98 (0.92, 1.05)  0.63

The nopvalue option suppresses p-value columns:

. regtab, coef("OR") noint nopvalue display

                              Model
───────────────────────────────────────────────────────
                              OR         95% CI
 Age at cohort entry (years)  1.00       (1.00, 1.00)
 Female sex                   0.99       (0.93, 1.06)
 Education level
   Primary                    Reference
   Secondary                  1.02       (0.94, 1.11)
   Tertiary                   1.09       (1.00, 1.18)
 Diabetes                     1.01       (0.94, 1.08)
 Hypertension                 1.10       (1.03, 1.17)
 Anxiety disorder             1.00       (0.93, 1.08)
 Prior cardiovascular disease 0.98       (0.92, 1.05)

Multinomial models keep outcome-specific rows and use RRR by default:

. collect clear
. collect: mlogit education index_age female diabetes hypertension, baseoutcome(1)
. regtab, display stats(n ll aic bic r2)

                                              Model
                                                RRR         95% CI   p-value
 Secondary: Age at cohort entry (years)        1.00   (1.00, 1.00)      0.75
 Secondary: Female sex                         1.08   (1.00, 1.18)     0.063
 Tertiary: Age at cohort entry (years)         1.00   (1.00, 1.00)      0.76
 Tertiary: Female sex                          1.02   (0.93, 1.11)      0.71
 Observations                                15,000
 AIC                                      32530.06
 BIC                                      32606.21
 Log-likelihood                         -16255.03
 Pseudo R²                                   0.000

Zero-inflated models keep the count and inflation equations distinct:

. collect clear
. collect: zip event_count treatment age_z female, inflate(zero_risk female)
. collect: zinb event_count treatment age_z female, inflate(zero_risk female)
. regtab, display stats(n aic bic ll) models("ZIP" \ "ZINB")

                                                 ZIP                                  ZINB
                                               Coef.           95% CI   p-value      Coef.           95% CI   p-value
 Event count: Treatment                        -0.15   (-0.26, -0.05)     0.003      -0.14   (-0.26, -0.03)     0.016
 Event count: Age z-score                       0.39     (0.34, 0.44)    <0.001       0.39     (0.33, 0.45)    <0.001
 Event count: Female                            0.30     (0.19, 0.41)    <0.001       0.32     (0.19, 0.45)    <0.001
 Inflation equation: Female                     0.50     (0.15, 0.84)     0.005       0.72     (0.26, 1.19)     0.002
 Inflation equation: Structural-zero risk       0.85     (0.66, 1.04)    <0.001       1.07     (0.79, 1.35)    <0.001
 Observations                                  1,500                                 1,500
 AIC                                        4338.98                               4307.53
 BIC                                        4376.17                               4350.04
 Log-likelihood                           -2162.49                              -2145.76

Hurdle models preserve outcome and selection equations while ancillary rows stay hidden in default presentation tables:

. collect clear
. collect: churdle linear annual_cost dose_intensity, select(participation_score) ll(0)
. regtab, display stats(n ll aic bic r2)

                                              Model
                                              Coef.         95% CI   p-value
 Annual cost: Dose intensity                   1.75   (1.49, 2.02)    <0.001
 Selection equation: Participation score       0.51   (0.43, 0.60)    <0.001
 Observations                                1,200
 AIC                                       3425.86
 BIC                                       3451.31
 Log-likelihood                          -1707.93
 Pseudo R²                                  0.100

corrtab — Correlation matrix

. corrtab index_age crp prior_hosp, ///
>     star(0.05 0.01 0.001) display

                             Age at cohort entry (years)  C-reactive protein (mg/L)  Prior hospitalizations
─────────────────────────────────────────────────────────────────────────────────────────────────────────────
 Age at cohort entry (years) 1.00
 C-reactive protein (mg/L)   -0.01                        1.00
 Prior hospitalizations      -0.00                        0.01                       1.00

* p<.05, ** p<.01, *** p<.001

crosstab — Cross-tabulation

. crosstab treated female, or label display

 Treatment group                          Male           Female         Total
────────────────────────────────────────────────────────────────────────────────
 SSRI                                     3,583 (59.4%)  5,351 (59.6%)  8,934
 SNRI                                     2,445 (40.6%)  3,621 (40.4%)  6,066
 Total                                    6,028          8,972          15,000
 Pearson's chi-squared test: chi2 = 0.06, p = 0.805
 OR = 1.0 (95% CI: 0.9, 1.1)

diagtab — Diagnostic accuracy

. diagtab phat cv_event, cutoff(0.35) auc wilson display

                Gold +    Gold -
──────────────────────────────────────────
 Test +         2,622     4,497
 Test -         2,627     5,254

 Measure        Estimate  (95% CI)
 Sensitivity    50.0%     (48.6, 51.3)
 Specificity    53.9%     (52.9, 54.9)
 PPV            36.8%     (35.7, 38.0)
 NPV            66.7%     (65.6, 67.7)
 Accuracy       52.5%     (51.7, 53.3)
 LR+            1.1       (1.0, 1.1)
 LR-            0.93      (0.90, 0.96)
 DOR            1.2       (1.1, 1.2)
 AUC            0.520     (0.510, 0.530)
 Youden's index 0.038

puttab + stacktab — emit-then-assemble export pipeline

puttab styles a table already in memory (the current dataset, a named frame(), or a Stata matrix()) as one house-styled sheet; stacktab imports those sheets and assembles them into a composite. Here two estimate/CI blocks are emitted with puttab, then stacked, column-merged, and section-labeled with stacktab:

. puttab term ahr ci using parts.xlsx, sheet("Block Primary") varlabels
puttab: wrote 3 data rows x 3 cols (data source) to sheet Block Primary in parts.xlsx

. puttab term ahr ci using parts.xlsx, sheet("Block Dose") varlabels
puttab: wrote 2 data rows x 3 cols (data source) to sheet Block Dose in parts.xlsx

. stacktab using parts.xlsx, sheet("Composite")            ///
      blocks(sheet(Block Primary) rows(1/4) cols(A-C) label(Any HRT use) \  ///
             sheet(Block Dose) rows(1/3) cols(A-C) label(By estrogen dose)) ///
      columnmerge(B+C as "aHR (95% CI)") spacing(1) display   ///
      title("Hormone therapy and recurrent events")           ///
      note("aHR = adjusted hazard ratio; CI = confidence interval.")

  ┌──────────────────────────────────────┐
  │      Any HRT use        aHR (95% CI)  │
  │          Any HRT   0.82 (0.69, 0.98)  │
  │    Former smoker   1.14 (0.97, 1.34)  │
  │   Current smoker   1.46 (1.21, 1.77)  │
  │                                       │
  │ By estrogen dose          aHR  95% CI │
  │         Low dose   0.91 (0.74, 1.12)  │
  │        High dose   0.73 (0.58, 0.92)  │
  └──────────────────────────────────────┘
stacktab: 2 blocks -> 8 rows written -> sheet Composite

The same pipeline writes the formatted workbook (title to A1, table from B2, merged aHR (95% CI) column, section dividers, and an italic note).

Persistent defaults

. tabtools set font Calibri
tabtools: default font set to Calibri

. tabtools set fontsize 11
tabtools: default font size set to 11

. tabtools set borderstyle thin
tabtools: default border style set to thin

. tabtools get
──────────────────────────────────────────────────
tabtools - Persistent Formatting Defaults
──────────────────────────────────────────────────

  Font:        Calibri
  Font size:   11
  Border:      thin

. tabtools set clear
tabtools: all persistent defaults cleared

Excel workbooks

The demo generates 14 workbooks (72 sheets) covering every command and option combination:

Integration with eplot: tables to forest plots

The same effect estimates that fill a publication table can drive a forest plot — without re-entering a single number. regtab and effecttab accept eplotframe(), which stores a graph-ready companion frame (label, estimate, ll, ul, pvalue, rowtype) that the separate eplot package reads directly. comptab and hrcomptab compose those companions and draw the plot in one step with forest.

The integration demo demo/demo_tabtools_eplot.do builds an adjusted odds-ratio table and turns it into a forest plot two ways. Run it from a local checkout:

stata-mp -b do tabtools/demo/demo_tabtools_eplot.do

One model: regtab table, then eplot forest

regtab writes the table and the companion frame at once; eplot plots the frame.

collect clear
quietly collect: logistic cv_event treated index_age female diabetes hypertension prior_cvd
regtab, coef("OR") noint eplotframe(or_effects, replace) display

eplot, frame(or_effects) labels(label) rowtype(rowtype) ///
    null(1) values stars vformat(%4.2f) ///
    effect("Odds Ratio (95% CI)") ///
    title("Predictors of cardiovascular events")

The table:

  +-----------------------------------------------+
  |                Model                          |
  |                   OR         95% CI   p-value |
  |      Treated    1.03   (0.96, 1.10)      0.38 |
  | Age at index    1.00   (1.00, 1.01)     0.035 |
  |   Female sex    1.01   (0.94, 1.08)      0.84 |
  |-----------------------------------------------|
  |     Diabetes    1.11   (1.04, 1.19)     0.002 |
  | Hypertension    1.01   (0.94, 1.08)      0.84 |
  |    Prior CVD    1.04   (0.97, 1.12)      0.23 |
  +-----------------------------------------------+

The matching forest plot:

Several models: comptab forest in one step

comptab composes companion frames from multiple regtab runs; forest calls eplot for you, and eplotoptions() passes graph options through.

* Crude and adjusted treatment effects, each captured as a regtab frame
collect clear
quietly collect: logistic cv_event treated
regtab, coef("OR") noint frame(g_crude, replace) eplotframe(ge_crude, replace)

collect clear
quietly collect: logistic cv_event treated index_age female diabetes hypertension prior_cvd
regtab, coef("OR") noint frame(g_adj, replace) eplotframe(ge_adj, replace)

comptab g_crude g_adj, rows(1 \ 1) section("Crude" \ "Adjusted") ///
    forest eplotoptions(null(1) title("Treatment effect: crude vs adjusted"))

Resources

• help tabtools for the suite overview and persistent defaults
• tabtools_tips or help tabtools_tips for compact option patterns and longer end-to-end recipes
• help tabtools_cheatsheet and help tabtools_cookbook as legacy aliases to tabtools_tips
• help table1_tc, help desctab, help regtab, help effecttab, help comptab, help hrcomptab, help survtab, help stratetab, help crosstab, help corrtab, help diagtab, help puttab, and help stacktab for command-specific syntax

Version History

• 1.6.4 (2026-06-10): Remove the retired workbook-composition alias from the shipped package. stacktab is now the only public command for assembling exported workbook blocks, and QA no longer installs or calls the old alias path.
• 1.6.3 (2026-06-10): Add tabtools_tips, a merged quick-reference and cookbook command/help topic. The former tabtools_cheatsheet and tabtools_cookbook help files are retained as compatibility aliases. Update the visible help and README command inventory for recently added commands including puttab, stacktab, and simtab.
• 1.6.2 (2026-06-08): Fix regtab AIC/BIC for GEE models. regtab reads model fit statistics from the active e(), but Stata's glm (the backend used by iivw_fit and xtgee-style GEE workflows) stores e(aic) as AIC per observation (AIC/N) and e(bic) under a deviance-based convention — neither comparable to the likelihood-scale values mixed and other ML estimators report. A stats(aic) table mixing GEE and mixed models therefore showed GEE AIC values roughly N times too small. regtab now always recomputes AIC as -2*ll + 2*k and BIC as -2*ll + k*ln(N) from e(ll), e(rank), and e(N) whenever those are available, overriding the per-observation/deviance-based stored values. Results now match estat ic for every model family and stay on one scale across rows. mixed output is unchanged (it already used this path). Added regression test qa/regtab/test_regtab_aic_gee.do.
• 1.6.1 (2026-06-08): Adds disk-backed tabtools defaults profiles. tabtools set ... , permanent writes the current house style to a runnable Stata do-file, using tabtools_profile.do in the PERSONAL ado directory by default or profile(filename) for project-specific profiles. tabtools use [using filename] reloads a saved profile into the current session, making fonts, themes, borders, digits, bold-p thresholds, and custom colors reproducible across sessions and projects. Refreshes table1_tc display defaults toward a compact house style: continuous variables now use format(%2.0f), percentages percformat(%5.0f) with no percent sign (percsign("")), SDs render as mean±SD (sdleft("±") sdright("")), IQRs as (Q1, Q3) (iqrmiddle(", ")), and low percentages are reported without a leading alignment space (no-space is now the default — the old behavior is available via the new spacelowpercent option, replacing nospacelowpercent). borderstyle(thin) is the default Excel border for both table1_tc and regtab, and regtab defaults to digits(2).
• 1.6.0 (2026-06-07): New command simtab — a Monte Carlo simulation performance table and export layer. Compute mode summarizes long replication-level results into table-grade measures (mean, bias, pctbias, empse, meanse, relerr, mse, rmse, coverage, power, n, nonconv) with closed-form Monte Carlo SEs used to flag off-nominal coverage; ingest mode (from(simsum)/from(siman)/from(summary)) renders an already-computed summary without recomputation, following the optional-dependency pattern used by comptab/hrcomptab with eplot. Multi-estimand tables get merged Excel group headers and flattened Markdown/CSV headers; nsim() adds non-convergence reporting; plotframe() provides a numeric figure companion. Cross-validated to exact agreement with simsum on bias/empirical SE/coverage and their Monte Carlo SEs. Pairs with simsum (White, Stata Journal 2010) and siman (UCL); cites Morris, White & Crowther (Stat Med 2019). Adds _simtab_ingest.ado, simtab.sthlp, and qa/simtab/.
• 1.5.2 (2026-06-06): Cleaner forest plots from comptab and hrcomptab. When a section() (or stratetab scaffold section) contributes exactly one plotted row, the eplot companion frame now folds the section label into that single row instead of emitting a standalone header row followed by one indented effect — the redundant header/child pair that made one-coefficient-per-model forests look cluttered. The rendered Excel and console tables are unchanged; only the eplotframe()/forest output differs. Added qa/_package/test_eplot_section_fold.do.
• 1.5.1 (2026-06-06): Fix two correctness bugs found while auditing the v1.5.0 eplot bridge. comptab and hrcomptab could not export Markdown (markdown() failed with rc=198 because of a malformed compound quote in the post-forest return block). regtab double-exponentiated logit, or and ologit, or models (the logit/ologit branch hardcoded eform=1 instead of respecting a user-supplied or option, unlike melogit/poisson/mlogit), silently reporting exp(OR); this also propagated into the eplot companion frame. Added regression tests for both in qa/_package/test_markdown_exports.do and qa/regtab/test_regtab_model_families.do.
• 1.5.0 (2026-06-06): Add an eplot bridge for graph-ready estimate/CI companion frames. regtab and effecttab now support eplotframe(); comptab and hrcomptab can compose those companions and draw forest plots with forest, passing graph options through eplotoptions() while honoring the active graph scheme by default. regtab and effecttab now support eplotframe(); comptab and hrcomptab can compose those companions and draw forest plots with forest, passing graph options through eplotoptions() while honoring the active graph scheme by default.
• 1.4.0 (2026-06-05): Add markdown() and mdappend exports across tabtools table commands, including same-call Excel plus Markdown export and sequential Markdown report building. Add _tabtools_markdown_write_current.ado as the shared Markdown writer and allow puttab to run Markdown-only without using.
• 1.3.7 (2026-06-03): Cap the label (first) column width in regtab, effecttab, and comptab so a single verbose row label — most commonly an unstructured random-effects Covariance: ... (slope, Intercept) row from a mixed model — can no longer stretch the whole column to 60-76 characters and balloon the table. The label column now caps at 45 characters by default; labels longer than the cap wrap onto extra lines (top-aligned) instead of being clipped by the adjacent estimate cell. The cap is tunable via the new labelwidth() option on all three commands.
• 1.3.6 (2026-06-01): Add puttab, a first-mile styled-block producer that writes a table already in memory — the current dataset, a named frame(), or a Stata matrix() such as e(b), r(table), or collapse/tabulate output — as one house-styled Excel sheet with the shared title/header/zebra/border geometry. For a matrix source the row and column names become the label column and header row; for a dataset or frame source numeric columns honor digits(), integers stay integer, and value labels are resolved. Repeated calls build a multi-sheet workbook that stacktab can assemble, closing the raw-input gap between desctab (needs a collect) and stacktab (needs pre-exported sheets). Add stacktab for block assembly of composite sheets. Together puttab and stacktab form the emit-then-assemble export pipeline.
• 1.3.5 (2026-06-01): Fix effecttab, digits() so collect-rendered 95% CI bounds use the requested decimal precision. Add regtab, cutlabels() for ordered-model cutpoints, make noint hide cutpoint and ancillary-only rows such as lnalpha, alpha, and /sigma, and split model-family demos into demo_regtab_models.xlsx with richer zero-inflated examples.
• 1.3.4 (2026-06-01): Extend regtab multi-equation row handling to zero-inflated Poisson, zero-inflated negative binomial, and Cragg hurdle models, with equation labels for outcome, inflation, selection, scale, and ancillary rows. Expand QA and demos for the model families covered by the regression-family matrix.
• 1.3.3 (2026-05-31): Make regtab preserve multi-equation row identity for estimators such as mlogit, auto-display multinomial logit output as relative risk ratios (RRR), and add a regression-family QA matrix covering mlogit, OLS, logit, probit, ologit, count, GLM, panel, survival, and quantile models.
• 1.3.2 (2026-05-29): Make regtab, stats() accept n_sub and subjects as synonyms for n (the N row, which already reports subjects for survival models), and warn instead of silently ignoring unrecognized stats() tokens.
• 1.3.1 (2026-05-27): Make regtab, relabel random-effect rows identify variance and covariance parameters explicitly, including linear mixed-model random-slope covariance rows such as cov(months_since_tx,_cons).
• 1.3.0 (2026-05-23): Replace final Excel writers with a shared Mata xl() backend, add Mata workbook read/write helpers for collect parsing and backend contracts, and remove export excel/import excel from command implementations.
• 1.2.0 (2026-05-20): Add regtab, nopvalue to suppress p-value columns from console, frame, CSV, and Excel and Markdown outputs while preserving internal p-values for significance stars and row highlighting.
• 1.1.0 (2026-05-13): Add desctab, a formatter for active table collections with per-statistic number formats, events / N (%) and other composite cells, Excel/CSV/frame/display outputs, and shared tabtools styling defaults.
• 1.0.15 (2026-05-07): Fix regtab ICC cross-pollution where a multi-model collection ending in mepoisson/menbreg silently suppressed ICC for all earlier mixed-effects models (now skipped per-model). Strip thousands separators from coefficient and CI cells so digits(), stars, boldp, dimnonsig, and r(table) work for coefficients ≥ 1000. Make reference-category detection match the underlying numeric value (0 or 1 with empty CI) instead of the rendered string, so non-default precision still labels rows "Reference". Emit a noisily warning when the per-model stats fallback fires for a multi-model collection. Document table1_tc reserved by() variable name prefixes (N_, m_, _c…). Plug a Mata workspace leak in table1_tc Excel error path. Use a tempname instead of the literal beatles value-label fallback. Replace ad-hoc …2 suffix scratch columns in headerperc with tempvars to avoid name collisions. Move integer check before recast long, force to prevent silent truncation. Sthlp boldp colon-position fix.
• 1.0.14 (2026-05-05): Add QIC (Quasi-likelihood Information Criterion) support to regtab for GEE models. When stats(aic) is requested after xtgee, QIC is automatically computed and displayed since AIC is undefined for quasi-likelihood estimators. QIC can also be requested explicitly via stats(qic).
• 1.0.13 (2026-04-27): Documentation improvements across all .sthlp files and README. Enhanced corrtab, survtab, and diagtab help files with richer descriptions, additional examples, and "Also see" sections. Standardized author blocks with mailto links. Added {vieweralsosee} links to the cheatsheet.
• 1.0.12 (2026-04-27): Fix crosstab, or rr rd for 2x2 variables coded with nonzero category values by internally recoding observed levels to 0/1 before calling Stata's cc/cs; reject undefined requested association measures instead of silently omitting them; and validate table1_tc wt() and numeric by() values within the analysis sample so excluded rows do not trigger false hard failures.
• 1.0.11 (2026-04-26): Fix table1_tc, wt() smd weighted SMD calculations for continuous, categorical, and binary variables; fix headerperc with total(before|after); and document active collect side effects for regtab and effecttab.
• 1.0.10 (2026-04-26): Fix weighted crosstab, trend, enforce unique truncated stratetab matrix row names, hard-fail missing final effecttab workbooks, reject binary diagtab, optimal, add corrtab shape conflict checks, clarify cookbook runnable versus illustrative recipes, and strengthen QA/install isolation.
• 1.0.9 (2026-04-23): Fix regtab exporting a spurious blank trailing column. The _re_group_label internal variable was not being dropped before export because it was bundled in a capture drop with _ci_seen, which only exists under dimnonsig.
• 1.0.8 (2026-04-22): Clarity audit release with hardened export-return behavior, synchronized package metadata, and expanded QA around release gates and export failures.
• 1.0.7 (2026-04-18): Stata-Tools suite release covering direct descriptive tables, collect-based model formatters, file-based rate workflows, and frame-based composite builders.
• 1.0.6 (2026-04-17): Incremental refinement release during the Stata-Tools packaging cycle.
• 1.0.5 (2026-04-17): Incremental refinement release during the Stata-Tools packaging cycle.
• 1.0.4 (2026-04-16): Early public packaging milestone for the tabtools suite.

Author

Timothy P Copeland, Karolinska Institutet