On variable selection in joint modeling of mean and dispersion

The joint modeling of mean and dispersion (JMMD) provides an efficient method to obtain useful models for the dispersion, especially in problems robust design experiments. However, literature on JMMD there are few works dedicated variable selection this theme is still a challenge. In article, we propose procedure selecting variables JMMD, based hypothesis testing quality model’s fit. A criterion checking goodness fit used, each iteration process, as filter choosing terms that will be evaluated by test. Three types criteria were considered model our procedure. used were: extended Akaike information criterion, corrected specific proposed us, type adjusted coefficient determination. Simulation studies carried out verify efficiency all situations considered, proved effective quite satisfactory. process was applied real example from industrial experiment.