1. (StataNow) INTERVAL-CENSORED MULTIPLE-EVENT COX MODEL. Interval-censored multiple-event data commonly arise in longitudinal studies because each study subject may experience several types of events and those events are not observed directly but are known only to occur within some time interval. With new estimation command stmgintcox, you can fit a marginal Cox proportional hazards model to such data and account for possible dependence between event times for multiple events. This command can fit models to either single- or multiple-record-per-event data, and it supports time-varying covariates for all events or specific ones. stmgintcox also offers flexible ways to specify models with event-specific covariates.

After estimation with stmgintcox, you can create plots of covariate-adjusted survivor and other functions with stcurve and produce goodness-of-fit plots with estat gofplot. You can also generate event specific predictions; with predict’s new option event(), you can select the event for which you want predictions. You can use new postestimation command estat common to estimate and test average effects of covariates across all events.

2. (StataNow) CONTROL-FUNCTION LINEAR AND PROBIT MODELS. You can now fit control-function linear and probit regression models with the new cfregress and cfprobit commands. Control-function regression provides a flexible alternative to traditional instrumental-variable methods for models with endogenous variables. Include continuous, binary, fractional, and count endogenous variables. Flexibly incorporate interactions of endogenous variables. And easily test for endogeneity.
3. (StataNow) Do-file Editor current word and selection highlighting. The Do-file Editor will now highlight all case-insensitive occurrences of the current word under the cursor. It will also separately highlight all case-sensitive occurrences of the current selection. These features can be disabled in the Do-file Editor preferences dialog.
4. (StataNow) estat gofplot has new options for customizing goodness-of-fit plots. New graph options plot#opts() and byplot#opts() control the rendition of plots. After stmgintcox, new options events() and sepevents produce event-specific cumulative hazard functions; new option event#opts() controls the rendition of event-specific plots; and new options graph#opts() and by#opts() control the look of the #th graph.
5. (StataNow) stcurve has new options for customizing graphs of survivor and related functions. New graph options plot#opts() and atplot#opts() control the rendition of plots. After stmgitncox, new options events() and sepevents produce event-specific functions; new option event#opts() controls the rendition of event-specific plots; new option graph#opts() controls the look of the #th graph; and new option byopts() controls how subgraphs are combined and labeled.
6. graph export, when exporting to an SVG file, now has new option pre(), which determines how to handle sequences of spaces in text elements. By default, browsers and applications that support SVG files will collapse sequences of whitespace in a text element into a single whitespace. Specifying pre(on) will set the style for text elements to preserve sequences of whitespace. Specifying pre(off) will take no action on text elements that contain sequences of whitespace. The default is pre(off). See svg_options for more information.
7. margins after menl is faster (often up to 10 times faster).
8. icd10cm and icd10pcs have been updated for the 2025 fiscal year. Type icd10cm query or icd10pcs query to see information about the changes.
9. Stata’s PDF documentation is now produced by using LaTeX instead of TeX. Slight differences in the PDF files include improvements in page layout and a change in font. In addition, file sizes of the new documentation are smaller than those of the previous format. The new PDF documentation maintains all functionality and content of the previous PDF documentation.

© Copyright 1996–2026 StataCorp LLC. All rights reserved.

10. collect layout has the following improvements:
    1. collect layout now obeys string formats, as specified in option sformat() of command collect style cell, applied to results _r_lb and _r_ub when constructing the built-in composite result _r_ci.
    2. collect layout now obeys string formats, as specified in option sformat() of command collect style cell, applied to results _r_crlb and _r_crub when constructing the built-in composite result _r_cri.
11. graph option colorvar() did not allow temporary variable names. This has been fixed.
12. graph with suboption valuelabel of option label(), when the variable had value labels of the form “aa” “bb” (to produce multiline graph axis labels), produced extra space between the axis and the axis title. This has been fixed.
13. graph bar, graph box, graph dot, graph hbar, and graph hbox with option by() and either suboption relabel or sort of option over() did not apply the if condition to limit the axis range. This has been fixed.
14. margins is no longer allowed after menl if any expression uses the lag of an expression, L.{name:}, because the predictions involved are recursive.
15. predict with option csnell or mgale, when used after stintcox with if and option tvc() with multiple-record interval-censored data, produced an uninformative error message. This has been fixed.
16. regress, when specified with aweights, left out a constant when computing the log-likelihood values posted to scalars e(ll) and e(ll_0). The omitted constant is half the sum of the natural log of the normalized weights. This omission did not affect the test statistic or p-value reported by lrtest. However, if you compared the log-likelihood value from regress with aweights with an equivalent model fit via maximum likelihood, such as intreg or gsem, the difference in the reported values was this constant. This omission also affected the reported values of AIC and BIC in estat ic and estimates stats. This has been fixed.
17. sem, when specified with latent variables in option means() and when one of those latent variables X has a path to another latent variable Y, unintentionally defaulted to including the intercept for latent variable Y. This behavior was due to a change in the Stata 17 update on 13dec2021 (item 3). Prior to this change, latent endogenous variables like Y would include only an intercept if you specified a path from _cons. The original behavior has been restored.
18. stintcox has the following fixes:
    • stintcox’s display options emlog(), emdots, and vcedots() produced an error message and did not override the general nolog option. This has been fixed.
    • stintcox with a multiple-record-per-subject dataset, when the status variable used in option status() contained missing values, exited with an error message instead of proceeding by excluding the missing observations. This has been fixed.

19. table, when specified with rowvar or colvar with a value label attached to extended missing value .m and option missing was not specified, used the value label attached to .m for the variable’s marginal total instead of Total. Similarly, if option missing was specified but .m was not present in the dataset, the value label attached to .m was used for the marginal total instead of Total. This has been fixed.
20. testnl and nlcom, when working with estimation results that have spaces in one or more equation names, exited with error message “conformability error” even when the expression was correctly specified. This has been fixed.