Risks of Divorce: Comparison between Cox and Parametric Models
​
Sanizah Ahmad, Norin Rahayu Shamsuddin, Nur Niswah Naslina Azid @ Maarof, and Hasfariza Farizad
​
​
Abstract. Divorce or also known as the dissolution of marriage occurs when the bond of matrimony between married couples is dissolved. Since the rate of divorce is on the rise all around the world, this study aims to identify potential risk factors contributing to divorce by making comparisons between Cox Proportional Hazards (PH) model (a semi-parametric method) with Weibull and Lognormal models (parametric methods) using survival data. We retrospectively studied 531 secondary data of the Muslim couples who filed for divorce in Selangor, Malaysia. The age at marriage of husband and wife, the presence of children, duration of marriage, couples' educational level and employment status, household income and counseling session were identified as potential risk factors. The AIC (Akaike Information Criterion) were used to compare the efficiency of models between the three methods. The Cox PH model gives the best fit with respect to the lower AIC value. The survival result from the Cox model showed that age at marriage of husband and attending counseling session significantly affect the decision to divorce.