AUTOMATIC MODEL BUILDING FOR BINARY LOGISTIC REGRESSION BY USING SPSS 20 SOFTWARE

摘要

This article describes the main effects of automatic logistic regression model building by the computer software. To illustrate different techniques which are available for the automatic model building of binary logistic regression the small firms' internationalisation grow data analysis was chosen and SPSS 20 software, which encloses a wide array of services like data management foresee accurate case-police statistics and different tests alongside their predictions. The model building investigations refer to continuous internationalisation process that consequently leads to the commitment of the firms to international markets. In this reason the best-subset search procedures such as forward stepwise, backward stepwise, forward entries, backward removal, were used to help to identify some of the most significant factors, influencing the development processes of internationalisation of high growth firms (HGFs). Stepwise logistic regression methods, specifically the forward stepwise and backward stepwise methods, were used to perform a stepwise selection of predictor variables. All effects of these automatically built models were evaluated by Hosmer-Lemeshow goodness-of-fit test, deviance, Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC) and other tests. Furthermore, the ROC curve analysis was used to measure the goodness-of-fit and to compare the built competing logistic regression models.