Current Issue

Estimating Longevity Using Non Parametric and Semi Parametric Survival Functions

Ajayi Moses Adedapo1­, Agbona Anthony Adisa2, Masopa Adekunle Nurudeen3

 DOI : https://dx.doi.org/10.31142/rajar/v4i6.04

 Author's Affiliations

1,2,3Department of Statistics, The Federal Polytechnic, Ede


 

ABSTRACT

The conventional models that are frequently used to summaries the general health of people in different nations are longevity and life expectancy. There are many other different methods of estimating life expectancy and these different methods give widely different answers. When selecting a method for estimating life expectancy, it is important to ensure that the method used is suitable for the data available and for the life history of the respondents. Although there is rarely only one correct method of summarizing demographic information, the problem of life table is that follow cohorts for long periods of time are not common, which prevents cohort analysis, and the critical assumption of a stable-age distribution so difficult to meet. Our derived longevity survival based models form parametric and non-parametric clearly demonstrated that the conventional life table and survival methods are clearly inconsistent and give misleading results. This study utilised University academic retirees data obtained from two premier Universities in Western Nigeria. The estimated mean life expectancy from life table model of Universities academic retirees is 18 years and estimated mean of post retirement years for Universities academic retirees from derived longevity using Kaplan Meier model is 22 years. Utilisation of explanatory variable from derived longevity using Cox proportional model estimated mean of post retirement years for universities academic retirees is 22 years.  Based on standard error estimate we can say that life table model is inappropriate for estimating life expectancy.