From: Challenges of using external data in clinical trials- an illustration in patients with COVID-19
Approach | Prior for\(\theta\) | Parameter | Mean | 95%CrI interval | |
---|---|---|---|---|---|
Ignoring external data | |||||
Simple Bayes | Beta(1,1) | \(\theta\) | 0.2407 | 0.1653 | 0.3253 |
Incorporating external data | |||||
Combining data | Beta(96,230) | \(\theta\) | 0.2801 | 0.2388 | 0.3233 |
Modifying the prior | |||||
- Based on quantiles | Beta(77,185) | \(\theta\) | 0.2771 | 0.2327 | 0.3239 |
- Based on shrinkage intensity,\(m=10\) | Beta(2.9,7.1) | \(\theta\) | 0.2408 | 0.0711 | 0.5927 |
- Based on shrinkage intensity,\(m=100\) | Beta(29.3,70.7) | \(\theta\) | 0.2637 | 0.2086 | 0.3856 |
Fixed Power Priors, EB estimate | \(\theta\) | 0.2725 | 0.2208 | 0.3240 | |
Random Power Priors | Beta(1,1) | \(\theta\) | 0.2479 | 0.1770 | 0.3176 |
\(a_0 \sim\)Beta(1,1) | 0.1955 | 0.0001 | 0.6212 | ||
\(\theta\) | 0.2390 | 0.1675 | 0.3192 | ||
\(a_0 \sim\)Beta(1/2,1/2) | 0.1072 | 0.0001 | 0.4887 | ||
\(\theta\) | 0.2473 | 0.1697 | 0.3160 | ||
\(a_0 \sim\)Exp(100) | 0.1930 | 0.0001 | 0.6232 | ||
Hierarchical Bayesian Models | |||||
Beta(1,1) | \(\theta\) | 0.2944 | 0.2480 | 0.3457 | |
Beta(1/2,1/2) | \(\theta\) | 0.2934 | 0.2450 | 0.3440 | |
Pocock’s bias Approach | N(0, 0.03) | \(\delta _h\) | -0.00035 | -0.0629 | 0.0549 |
\(\theta\) | 0.2770 | 0.2355 | 0.3194 | ||
N(0.07, 0.03) | \(\delta _h\) | -0.0697 | 0.00706 | 0.1249 | |
\(\theta\) | 0.2770 | 0.2365 | 0.3194 | ||
\(N(-0.07,0.03)\) | \(\delta _h\) | -0.0703 | -0.13296 | -0.0151 | |
\(\theta\) | 0.2770 | 0.2365 | 0.3194 |