yesung000 发表于 2021-6-12 20:07
请问你做出来了吗,最近也在看这个模型,头大
The hyper prior for the precision (inverse variance) in the random walk priors is a Gamma distribution with parameters a and b;expected value is a/b, variance is a/b^2. Weak hyper parameters are suggested, defaults are a = 1, b = 0.0005 for age, period and cohort effects and a = 1, b =0.05 for overdispersion (if added).