Stochastic modeling of mortality in Two Population: application to limited data situation

Moyosola A. Bamidele, JHU/CCP Nigerian Urban Reproductive Health Initiative
Olusola Adejumo, University of Ilorin
Ruth Ella, University of Ilorin

The appropriateness of mortality models varies from country to country depending on the prevailing parameter of the country. This paper extended existing stochastic mortality model to capture mortality situation in two populations with application to limited data situation. Data were generated from binomial distribution using Monte-carlo simulations; the mortality rate for age 0 to 100years and 3-year points to capture the limited data condition and the model were applied to Nigeria mortality data. Bayesian Information Criterion (BIC), Deviance and Log-likelihood were used to assess the performance of the models. The confidence intervals of the forecast from the ARIMA models were very wide indicating that the future pattern of mortality is not stable but over time might come to equilibrium while the cohort effect estimate from the Nigeria was not significant. The findings from the study shows the appropriate of AP and ASPC model in modeling mortality in two-population situation.