XTSEL: Stata module for selection of variables and specification in a panel-data framework

The package XTSEL includes two new commands (xtselvar and xtselmod) that help us to rank the best predictors between a number of alternative explanatory variables (candidates), or the best specification between all possible combinations of a set of explanatory variables, according to several in-sample and out-of-sample statistics. They are specially adapted for a panel data framework, firstly because the out-of-sample prediction performance is measured in the two inherent dimensions of a panel (time-series and cross-individuals), and secondly because they allow a large number of methodological options that typically are necessary in panel data analysis. Given a set of n predictors, xtselvar estimates the same specification n times, one for each predictor. xtselmod estimates 2^n - 1 different specifications, one per each possible combination out of the set of n variables. Both procedures keep constant the same dependent variable and an optional list of fixed control variables, plus several other methodological options. For each candidate variable/specification, the procedures estimate a set of parameters and statistical criteria: 1. Adjusted R squared (R2_ad). 2. Akaike Information Criterion (AIC). 3. Bayesian Information Criterion (BIC) 4. U-Theil in time-series dimension: RMSE of variable/specification vs. RMSE from a naïve prediction or an AR1 model (Uth_TS). 5. U-Theil in cross-section dimension: RMSE of variable/specification vs. RMSE from a naïve prediction or an AR1 model (Uth_CS). Both commands rank each variable/specification according to each criterion and generate one ranking per each one of them. xtselvar also reports coefficients and t-statistic of each candidate variable. They also compute a composite ranking summarizing all five criteria. They finally sort all candidate variables/specifications according to the selected ranking, which by default is the composite ranking.