Fitting Mixtures of Logit Regressions with the Forward Search

• Activity Number: 584
• Type: Contributed
• Date/Time: Wednesday, August 3, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Business and Economic Statistics Section
• Author(s): Margherita Velucchi*+ and Matilde Bini
• Companies: University of Florence and European University of Rome
• Keywords: Forward search ; Robustness ; Logit Regression ; Firms Behavior

Abstract

The forward search is a powerful general method to detect multiple outliers and to determine their effect on inferences about models fitted to data. From the monitoring of a series of statistics based on subsets of data of increasing size we obtain several views of any hidden structure. One of the problems of the forward search has always been the lack of an automatic link among the great variety of plots which are monitored. Sometimes, some features emerge unexpectedly during the progression of the forward search only when a specific combination of forward plots is inspected at the same time. These features have to be harmonized and linked together in order to give an exhaustive description of a very complex problem. In this paper, we aim at filling this gap and use a set of new robust graphical tools whose power will be demonstrated on a logit regression problem. We use real microeconomic data (Italian firms balance sheets in 2007) and simulated data and we show the dynamic interaction with different "robust plots" to highlight the presence of groups of outliers and regression mixtures and appraise the effect that these hidden groups exert on the fitted model.