Cohort size and maximum likelihood estimation of mortality parameters

The purpose of this study was to investigate the effects of cohort size on maximum likelihood estimates of mortality parameters. Recent experimental investigations have stressed the importance of large cohorts for detecting leveling off of mortality rates at older ages. In the present study, emphasis was placed on evaluation of relatively small cohorts (about 150-300 individuals). Deaths were simulated under the assumption of the frailty mortality model. Two different parameter sets that resulted in differences in mean life span of more than twofold were used for simulations. Our smallest cohorts yielded parameter estimates that had generally good statistical properties, but relatively large standard errors. For tests of hypotheses concerning equality of parameters among populations or experimental treatments, empirical standard errors (obtained from several cohorts) were preferable to asymptotic standard errors (obtained for single cohorts). In particular, empirical standard errors yielded reliable type I error rates.