Identifying the determinants of lifelong singlehood via boosted mixture cure models

Guillermo Briseño Sanchez, Karlsruhe Institute of Technology

Co-authors: Angela Carollo, Max Planck Institute for Demographic Research; Nadja Klein, Karlsruhe Institute of Technology

Abstract: We combine distributional mixture cure models and automatic variable selection via component-wise boosting to analyse romantic partnerships using demographic data from Germany. By defining lifelong singles as individuals who do not enter a co-residential relationship by age 50, we frame the problem as a mixture cure time-to-event model. Our approach allows to estimate the proportion of individuals who will remain single as well as to identify the most important predictors for the probability of remaining single, and entering a cohabitation.