sequential specification search

Sequential specification search is a backward model selection approach that starts with a full model which contains all possible independent variables and then systematically removes variables while checking if the removed variables still keep the reduced model statistically significant. The goal of this method is to find a parsimonious, but accurate model with the least amount of independent variables. This method is advantageous because it narrows the possible models and parameters quickly and efficiently, by sequentially removing those variables and parameters that have less effects on the dependent variable. Also, since all the models are created from the same base model, the same set of assumptions are applied to all the reduced models. This results in a more consistent estimation process compared to using a different base model for each reduced model.