The Stata Journal: Promoting communications on statistics and Stata

As explained in the Editorial announcements at the start of this issue, the Stata Journal will be featuring various interviews with selected members of the Stata community in 2010, Stata's 25th anniversary year. In this issue, we start with an online interview, dated October 14, 2009, of Christopher F. “Kit” Baum, an economics professor at Boston College. Kit has been highly active as a Stata user since the 1990s, contributing as an author and associate editor to this journal and its predecessor, the Stata Technical Bulletin; as a much-downloaded Stata program author; as an author of two notable Stata-based texts; as a frequent participant on Statalist and at Users Group meetings in both the United States and several European countries; and as founder and maintainer for more than a decade of the SSC archive, which now contains many hundred user-written Stata packages. In this interview, he comments on how he got into Stata and adds his own speculations for the future.

In this article, I discuss a method by Erikson et al. (2005, Proceedings of the National Academy of Science 102: 9730–9733) for decomposing a total effect in a logit model into direct and indirect effects. Moreover, I extend this method in three ways. First, in the original method the variable through which the indirect effect occurs is assumed to be normally distributed. In this article, the method is generalized by allowing this variable to have any distribution. Second, the original method did not provide standard errors for the estimates. In this article, the bootstrap is proposed as a method of providing those. Third, I show how to include control variables in this decomposition, which was not allowed in the original method. The original method and these extensions are implemented in the ldecomp command.

The World Bank's world development indicators (WDI) compilation is a rich and widely used database about the development of most economies in the world. However, after insheeting a WDI dataset, some data management is required prior to performing statistical analysis. In this article, I propose a new Stata command, wdireshape, for automating this data management. While reshaping a WDI dataset into structures amenable to panel data, seeming unrelated regression, or cross-sectional modeling, wdireshape renames the series and places the series descriptors into variable labels.

The SPost user package (Long and Freese, 2006, Regression Models for Categorical Dependent Variables Using Stata [Stata Press]) is a suite of postestimation commands to compute additional tests and effects representations for a variety of regression models. To facilitate and automate the task of tabulating results from SPost commands for inclusion in reports, publications, and presentations, we introduce tools to integrate SPost with the estout user package (Jann, 2005, Stata Journal 5: 288–308; 2007, Stata Journal 7: 227–244). The estadd command can retrieve results computed by the SPost commands brant, fitstat, listcoef, mlogtest, prchange, prvalue, and asprvalue. These results can then be tabulated by esttab or estout.

In this article, we describe a command, riskplot, aiming to provide a visual aid to assess the strength, importance, and consistency of risk factor effects. The plotted form is a dendrogram that branches out as it moves from left to right. It displays the mean of some score or the absolute risk of some outcome for a sample that is progressively disaggregated by a sequence of categorical risk factors. Examples of the application of the new command are drawn from the analysis of depression and fluid intelligence in a sample of elderly men and women.

We present a new Stata estimation program, mboxcox, that computes the normalizing scaled power transformations for a set of variables. The multivariate Box–Cox method (defined in Velilla, 1993, Statistics and Probability Letters 17: 259–263; used in Weisberg, 2005, Applied Linear Regression [Wiley]) is used to determine the transformations. We demonstrate using a generated example and a real dataset.

Availability of large, multilevel longitudinal databases in various fields including labor economics (with workers and firms observed over time) and education research (with students and teachers observed over time) has increased the application of panel-data models with multiple levels of fixed-effects. Existing software routines for fitting fixed-effects models were not designed for applications in which the primary interest is obtaining estimates of any of the fixed-effects parameters. Such routines typically report estimates of fixed effects relative to arbitrary holdout units. Contrasts to holdout units are not ideal in cases where the fixed-effects parameters are of interest because they can change capriciously, they do not correspond to the structural parameters that are typically of interest, and they are inappropriate for empirical Bayes (shrinkage) estimation. We develop an improved parameterization of fixed-effects models using sum-to-zero constraints that provides estimates of fixed effects relative to mean effects within well-defined reference groups (e.g., all firms of a given type or all teachers of a given grade) and provides standard errors for those estimates that are appropriate for shrinkage estimation. We implement our parameterization in a Stata routine called felsdvregdm by modifying the felsdvreg routine designed for fitting high-dimensional fixed-effects models. We demonstrate our routine with an example dataset from the Florida Education Data Warehouse.

The development and use of synthetic regression models has proven to assist statisticians in better understanding bias in data, as well as how to best interpret various statistics associated with a modeling situation. In this article, I present code that can be easily amended for the creation of synthetic binomial, count, and categorical response models. Parameters may be assigned to any number of predictors (which are shown as continuous, binary, or categorical), negative binomial heterogeneity parameters may be assigned, and the number of levels or cut points and values may be specified for ordered and unordered categorical response models. I also demonstrate how to introduce an offset into synthetic data and how to test synthetic models using Monte Carlo simulation. Finally, I introduce code for constructing a synthetic NB2-logit hurdle model.

Mata is Stata's matrix language. In the Mata Matters column, we show how Mata can be used interactively to solve problems and as a programming language to add new features to Stata. The subject of this column is using Mata to solve data analysis problems with the new Stata commands putmata and getmata, which were added to official Stata 11 in the update of 11 February 2010.

The statsby command collects statistics from a command yielding r-class or e-class results across groups of observations and yields a new reduced dataset. statsby is commonly used to graph such data in comparisons of groups; the subsets and total options of statsby are particularly useful in this regard. In this article, I give examples of using this approach to produce box plots and plots of confidence intervals.