Stata commands for multi-period experimental data

Multi-period experimental designs are a popular way to generate more observations and examine changes in participant behavior over time. However, choosing the most appropriate model to analyze the resulting data can be challenging. Two Stata commands, qic and xtgee, help us select among different empirical models.

Xtgee and QIC

The xtgee command fits generalized linear models and allows you to specify the distribution of the dependent variable, the link function, and, the correlation structure for multi-period data. Different combinations of these three options fit popular panel models in Stata (See figure below).
On its own the xtgee command already gives us substantial control over our model specification. Selecting the most appropriate model for our multi-period data, however, can be challenging. To select among different models estimated by xtgee, James Cui developed the qic command (see this article). The qic command calculates the QIC and QIC_u criteria for model selection in xtgee, which is an extension of the widely used AIC criterion in regressions (Pan, 2001). Lower QIC and QIC_u values mean a model is preferred over other models. Using the QIC and QIC_U criteria, Stata users can, therefore, evaluate which empirical model is most appropriate for their data.

How to reference this online article?

Van Pelt, V. F. J. (2019, September 2). Stata commands for multi-period experimental data. Accounting Experiments, Retrieved from: https://www.accountingexperiments.com/post/period-regressions/.