Smi-log plot, GINPPI mortality rate

The amazingly linear log(m)~t indicate m ~ exp(t), Gompertz model. Lifespan of 2000 individuals were simulated using merged GINPPI, p=0.7, lambda=0.005

I can also identify the best linear section based on R^2. The best linear section is informative on 'initial virtual age'.

Todo: overlay simulated lifespan with different lambda in p=0.7 figures. Esimate G and R and Makeham constant from the linear plot.


When lamba is small and p is high, the initial bin can become too small, as little as 1 per bin. This would lead to highly inaccurate mortality rate estimations. This numerical problem can be mitigated by increase simulation sample size.

The above simulation shows that when bin-size < 5 (in both left and right tails), the mortality rate estimation are very noises.