I am building a model an additive model that incorporates standard errors (SE) and regression weight, which are the sample sizes. The data I am using come from the literature, so the approach is meta-analytic. Some authors published their raw data, in order words, individual observations with sample size = 1 and standard error = 0. Some authors reported average +/- SE with sample size > 1.
The problem is that the family of distributions I am using is gaussian with a log link, and an error arises when I input SEs of 0. Here is the structure of my model and the output when I try to run it:
fit1 <- brm(response|weights(Weight) + se(SE) ~ X1 + X2 + X3..., family = gaussian(link="log"), data = Data)
Chain 1: Rejecting initial value: Chain 1: Error evaluating the log probability at the initial value. Chain 1: Exception: normal_lpdf: Scale parameter is 0, but must be > 0! (in 'model125839f2ad0_c2a2bf402f16cab2ec43c0c861aee5d3' at line 40)
I am wondering if there is some way to combine averages and individual samples in brms, or if anyone has suggestions on how to move forward?
Thank you for your time!