Modelling Rare Events Data: An application on the association between education and depression


Göktaş D.

VII. Anadolu International Conference on Economics (ECONANADOLU’24), Eskişehir, Turkey, 31 May - 01 June 2024, pp.1

  • Publication Type: Conference Paper / Summary Text
  • City: Eskişehir
  • Country: Turkey
  • Page Numbers: pp.1
  • Bilecik Şeyh Edebali University Affiliated: Yes

Abstract

The prevalence in general population is 1 per cent for severe depression, 2 per cent for moderately severe depression, as reported in the Turkish Health Survey 2022. These rare rates can exhibit   a separation problem resulting from the use of logit or probit estimates.

This study attempts to compare different generalized linear models to gauge possible the association between invidiual's educational attainment and having depression, using depression-related questions from the Türkiye Health Survey 2022. I carry out logit estimation, clog-log link function, and firth logistic regression considering their Akaike Information Criteria. It is anticipated that clog-log estimation will outperform the other two models.