We present the development and usage of the generalized model specification system (GMSS) Mata library to support the specification of regression models. This library allows users to gain access to all the options in Stata’s premier optimization routines. The library consists of classes for distributions and link functions. In addition, the library defines optimization routines callable from moptimize(). The library consists of approximately 30 distributions for which users may specify associated covariates for each of the parameters. General models (zero-inflated, zero-altered, zero-marginalized, zero-truncated, and heaped) are usable with any count distribution in the library. Users are also free to develop and add distributions and link functions. All distributions have default link functions, but users are free to specify links as well. An ado-file is available for those users who prefer Stata-language implementations. Support commands for estat and predict are also included in the library. Factor-variable specification of covariate lists is allowed, and the library automatically generates associated constraint matrices. Example usage of do-files calling the Mata routines will be presented along with usage of the ado-file specification. While there is substantial overlap with specific existing models, the GMSS library includes many new distributions for interested developers and applications.

Pre- and posttest visualizations are powerful—they can tell a story of change or lack of change within a population and help researchers inform their insights into humanity and the surrounding environment. Stata is very useful for gathering insights from pre- and posttests, and in conjunction with Excel and Tableau, one has the capacity to create pre- and posttest visualizations that are efficient and visually appealing.

To create these visualizations, I will discuss a multistep method. The use of Stata functions, such as asdoc and reshape, will be a critical component to the data analysis process. To accompany this, data manipulation and data functions, such as if-then statements, will be used in Excel from Stata output to generate a dataset that can be imported into the Tableau interface. And with the Tableau interface, visualizations can be created that can provide insights into sample sizes, statistical significance, and change all in one view.

Linear panel models and the “event-study plots” that often accompany them are popular tools for learning about policy effects. We introduce the xtevent package, which enables the construction of event-study plots following the suggestions in Freyaldenhoven et al. (Forthcoming). The package implements various procedures to estimate the underlying policy effects and allows for nonbinary policy variables and estimation adjusting for preevent trends.

Researchers often use control-function methods when traditional instrumental-variables methods lack desired flexibility. The new Stata commands cfregress and cfprobit allow for specification of control-function linear and probit models and accommodate continuous, binary, fractional, and count endogenous variables, all while returning appropriately adjusted standard errors. This presentation will give a practical introduction to control functions and show how these commands can be useful in empirical work.

If all you have are the original Stata regular expression functions in your toolkit, all text data can look like a nightmare. Even with better regular expression functions, the lack of high-quality documentation, guides, and examples makes it more difficult to use regular expressions effectively. Have you ever wondered about the differences between possessive and greedy expressions? Have you ever wondered how to use positive and negative look-ahead and look-behind functionality? Do you find some commands using regular expressions taking much longer to execute than you would like? If so, this is the talk for you.

During the talk, I will explain how regular expressions work, and what the different metacharacters, matching types, look-arounds, and different character classes do, describe how to improve the performance of your regular expressions, and describe misconceptions along the way.

Reproducibility is essential for understanding how social scientists reach their conclusions. Following recent reproducibility and transparency standards adopted by social science journals, ensuring that code is reproducible has become a priority for researchers and institutions. Reproducibility issues may arise from the use of commands that introduce unnoticed randomness, depend on system-specific settings, or create unexpected inconsistencies across code runs. To address this challenge, the Development Impact Analytics team at the World Bank has developed repscan, a Stata command designed to enhance the reproducibility of research code. Part of the repkit Stata package, repscan scans a do-file, detecting and flagging commands known to compromise reproducibility. By alerting users of potential problems—such as commands affected by uncontrolled randomness, system-dependent sortings, or unstable default behaviors—repscan allows researchers to refine their code and ensure their results can be consistently reproduced. This presentation will provide an overview of repscan’s functionality, demonstrating its application in typical coding tasks and showcasing how it can enhance reproducibility in social science research.

This study applies Stata’s quaidsce command to estimate a censored demand system for food consumption in Mexican households using data from the 2022 National Survey of Household Income and Expenditures. The high prevalence of zero expenditures across food groups presents a challenge for traditional demand estimation methods. quaidsce implements a two-step procedure that corrects selection bias, ensuring more accurate estimates of price and income elasticities. Our findings demonstrate that failing to account for censoring leads to systematic biases in elasticity estimates, with distortions increasing as the proportion of censored observations grows. By efficiently handling censored dependent variables, quaidsce enhances the reliability of demand system estimation, making it a valuable tool for researchers working with consumption and expenditure data.

Giordano and Broderick (2024) introduced infinitesimal jackknife (IJ) standard errors for Bayesian estimators (posterior means). Just like resampling standard errors, IJ standard errors are robust to model misspecification and can be adapted to account for clustering. Importantly, IJ standard errors do not require resampling but can be obtained from a single MCMC run. Standard Bayesian quantile regression, as implemented in bayes: qreg, is generally misspecified. This is because the motivation for the asymmetric Laplace (AL) likelihood is merely that its maximum coincides with the classical quantile regression estimator of Koenker and Bassett (1978). There is no reason to believe that the AL distribution is a plausible data-generating mechanism. For example, the shape of the distribution depends on the quantile you are interested in. While point estimation is consistent, credible intervals often have poor frequentist coverage. We therefore propose using IJ standard errors for Bayesian quantile regression and show, via simulations, that they have good frequentist properties, both for independent and clustered data. If made available as an option in bayes: and bayesmh, IJ standard errors may soon become as popular for Bayesian inference as the vce(robust) option for frequentist inference.

Giordano, R., and Broderick, T. 2024. The Bayesian infinitesimal jackknife for variance. arXiv:2305.06466.
Ji, Feng, Lee, JoonHo, and Rabe-Hesketh, S. 2024. Valid standard errors for Bayesian quantile regression with clustered and independent data. arXiv:2407.09772v1.

We propose a per-cluster instrumental-variables approach (PCIV) for estimating correlated random coefficient models in the presence of contemporaneous endogeneity and two-way fixed effects using Stata. Our estimator uses variation across clusters to estimate coefficients with homogeneous slopes (such as time effects) and within-cluster variation to estimate the cluster-specific heterogeneity. We aggregate cluster-specific estimates to population averages. We demonstrate consistency, showing robustness over standard estimators, and provide analytic standard errors for robust inference. Our Stata package allows for straightforward implementation. In Monte Carlo simulation, PCIV performs relatively well against pooled 2SLS and fixed-effects IV (FEIV) with a finite number of clusters or finite observations per cluster. We apply PCIV in estimating the price elasticity of gasoline demand using state fuel taxes as instrumental variables. PCIV estimation allows for greater transparency of the underlying data. It produces graphs depicting divergence in the implicit weighting when applying FEIV from the natural weights applied in PCIV and evidence of correlations between heterogeneity in the first and second stages, violating a key assumption underpinning the consistency of standard estimators. In our application, overlooking effect heterogeneity with standard estimators is consequential. Our estimated distribution of elasticities reveals significant heterogeneity and meaningful differences in estimated averages.

Cost-effectiveness analysis (CEA) is often conducted alongside a randomized clinical trial to establish whether the new therapy is likely to have a favorable value for its cost. One common approach is to estimate an incremental cost-effectiveness ratio (ICER), the marginal health benefit relative to the marginal cost, and compare the point estimate with a prespecified “willingness to pay”. Alternatively, net monetary benefit (NMB) may be used to keep benefits and costs linear and on the same scale. Costs and benefits may be modeled separately and often assume different distributions, especially for the ICER, where benefits are generally constrained to the -1 to 1 interval. CEA also involves use of graphs to assess uncertainty about the decision being made. The first, the cost-effectiveness plane, plots bootstrapped replicates of incremental cost against incremental benefits with confidence ellipses. The second, the cost-effectiveness acceptability curve (CEAC), is a plot showing the probability that a new treatment will be cost-effective at different willingness-to-pay values. In this presentation, we will introduce a new suite of Stata CEA commands. They use standard Stata command syntax to fit models and obtain the ICER or NMB and then provide comprehensive postestimation support and graphing.

Join us for a happy hour celebrating 40 years of Stata! All registered participants are welcome to attend.