Stata 18 May 2026

For high-dimensional control variable selection, pdslasso implements the Belloni, Chernozhukov, and Hansen (2014) method. It uses Lasso to select relevant controls from a large set and then performs valid inference on a single treatment variable—perfect for situations with more potential confounders than observations.

Before diving into the technical nuances, here is a high-level overview of what Stata 18 brings to the table:

Let’s explore each of these areas in detail. Stata 18

Stata has significantly strengthened its causal inference toolkit, a critical area for observational studies.

Yes, but computationally intensive tasks like Bayesian MCMC will be slow. For large projects, 16 GB of RAM and an SSD are strongly recommended. Before diving into the technical nuances, here is

bayes, prior(sd): melogit outcome x1 x2 || cluster: x1
* Now can use any prior, any sampler
bayesmh y x1 x2, likelihood(normal) prior(y:, flat) ...

Previously, you needed third-party tools for literate programming with Stata. Now, Stata 18 has built-in support for Stata Markdown (.smd files). You can write a document that interleaves narrative text, Stata code, and output (tables, graphs). One click executes the code and renders to HTML, PDF, or Word.

This is Stata’s answer to R Markdown and Jupyter notebooks, tailored for Stata users who want reproducibility without leaving the environment. Let’s explore each of these areas in detail

Stata 18, released by StataCorp in April 2023, represents a significant leap forward in statistical software, blending traditional robustness with cutting-edge analytical techniques. It builds on Stata’s renowned ease of use, reproducibility, and documentation while introducing major advancements in causal inference, Bayesian analysis, reporting, and data visualization.

Stata has been building its Bayesian capabilities for several releases, but Stata 18 makes Bayesian analysis accessible to the average researcher while adding power for the specialist.