Modelling gained university credits: mixtures vs quantile regression
Leonardo Grilli, Dept. of Statistics, Computer Science, Applications "G. Parenti", Univ. of Florence
Venerdì 27 novembre 2015, ore 12,00
Aula 202
Abstract:
The talk will consider two recent papers by Grilli, Rampichini and Varriale about modelling credits obtained by university freshmen during the first year, in order to investigate whether the pre-enrolment assessment test is an effective tool to predict student performance. Looking at data from the School of Economics of the University of Florence, it appears that gained credits is a count variable with an irregular distribution and a peak in zero. This pattern represents a challenge in statistical modelling, which is tackled using two distinct approaches: (i) a concomitant variable binomial mixture model, and (ii) a two-part model with a logit specification for the zeros, while positive values are analyzed by quantile regression for counts. The two approaches are applied to the same dataset, discussing issues of estimation and interpretation.