Stata 15 features Buy / upgrade now


Combine endogenous covariates, sample selection, and endogenous treatment in models for continuous, binary, ordered, and censored outcomes.

Latent class analysis (LCA)


Discover and understand the unobserved groupings in your data. Use LCA's model-based classification for find out:

  • How many groups you have;
  • Who is in those groups; and
  • What makes those groups distinct.

bayes: logistic and more

Type bayes: in front of any of 45 Stata estimation commands to fit a Bayesian regression model.

Linearised DSGEs


Write your model in simple algebraic form. Stata does the rest: solve model; estimate parameters; estimate policy and transition matrices (with CIs); estimate and graph IRFs; and perform forecasts.

Markdown and dynamic documents

  • Create webpages from Stata
  • Intermix text, regressions, results and graphs
  • See changes in data or commands automatically reflected on your webpages

Finite mixture models (FMMs)

  • 17 estimators and combinations
  • Continuous, binary, count, ordinal, categorical, censored, and truncated outcomes
  • Survival outcomes

Spatial autoregressive models

Because sometimes where you are matters.

  • Spatial lags of dependent and independent variables
  • Spatial lags of autoregressive errors

For example: Neighboring towns have more influence on each other than on towns far away. The same is true of countries that are close to each other and of closely connected friends on social media.

There is a new manual entirely devoted to fitting SAR models, working with spatial data, and creating and managing spatial weighting matrices. The new commands are called the Sp commands.

Interval censored survival models

Fit any of Stata's six parametric survival models to interval-censored data. All the usual survival features are support:

  • stratified estimation;
  • robust and clustered SEs;
  • survey data;
  • graphs;
  • and more.

Nonlinear mixed-effects models


When your sciences says your model is nonlinear in its parameters...

Stata now fits nonlinear mixed-effects models, also known as nonlinear multilevel models and nonlinear hierarchical models. These models can be thought of in two ways. You can think of them as nonlinear models containing random effects. Or you can think of them as linear mixed-effects models in which some or all fixed and random effects enter nonlinearly. However you think of them, the overall error distribution is assumed to be Gaussian.

Mixed logit models - Advanced choice modelling

  • Do you walk to work, ride a bus, or drive a car?
  • Which of three insurance plans do you buy?
  • Which political party do you vote for?

We make dozens of choices every day. Researchers have access to gaggles of data about those choices. Mixed logit introduces random effects into choice modelling and thereby relaxes the IIA assumption and increases model flexibility.

Nonparametric regression, kernel methods

Nonparametric regression

When you know something matters...

...but you have no idea how.

Create Word® and PDF documents from Stata

  • Automate your reports
  • Write paragraphs and tables to Word documents
  • Embed Stata results and graphs in paragraphs and tables
  • Customise formating of text, tables and cells

Threshold regression

Your time-series regression may change parameters at some point in time or at multiple points in time. The activity of foraging animals might follow a completely different pattern at temperatures above some threshold. You may not know the value of that threshold. Finding such threshhold and estimating the parameters within the regimes is what threshold regression does.

Stata/MP - more variables

Because speed is most important when our users’ problems are big, Stata/MP supports even larger datasets than Stata/SE. Stata/MP 15 now allows up to 120k variables, up from 32,767 with Stata/SE. Stata/MP still provides the most extensive multicore support of any statistics and data management package. Now that it supports larger datasets, more than ever, it is the clear choice for users who need both speed and size.

And MORE!!!!

  • Bayesian multilevel models
  • Panel-data tobit with random co-efficients
  • Multilevel regression for interval-censored data
  • Multilevel tobit regression for censored data
  • Panel-data cointegration tests
  • Tests for multiple breaks in time series
  • Multiple-group SEM for continuous, binary, ordered, and count outcomes
  • Multiple-group multilevel SEM
  • ICD-10-CM/PCS
  • Power analysis for cluster randomized designs
  • Power analysis for linear regression models
  • Heteroskedastic linear regression
  • Poisson with sample selection
  • Zero-inflated ordered probit
  • Add your own power and sample-size methods
  • Transparency on graphs
  • Stream random-number generator