Table 6 Estimation error of probability vector by VAM in simulation. The 95% CI is computed by sampling from identical distribution (Gamma(1, 3)) and compute the distances between the samples. The estimation error of population are listed in the column Sample. The estimation error of cohort are listed in the column Cohort. The estimation error based on the three different simulation dataset are listed in the last six rows
| Â | Â | Sample | Cohort |
|---|---|---|---|
95% CI in L infinity | [0.0012, 0.0068] | [0.00205, 0.0116] | |
95% CI in KL divergence | \([8.5*10^{-6}, 0.00019]\) | \([2.4*10^{-5}, 0.00056]\) | |
Simulation dataset 1 | L infinity | 0.00384 | 0.00085 |
KL divergence | 0.00007 | \(5*10^{-6}\) | |
Simulation dataset 2 | L infinity | 0.00384 | 0.00085 |
KL divergence | 0.00007 | \(5*10^{-6}\) | |
Simulation dataset 3 | L infinity | 0.00557 | 0.00490 |
KL divergence | 0.00024 | 0.00017 | |