lognormal_with_mean_parameter2

# based on the great vignette on specifying custom response distributions in brms:
# https://cran.r-project.org/web/packages/brms/vignettes/brms_customfamilies.html

library(languageR)
library(brms)

danish$RT <- exp(danish$LogRT) 
danish$cSex <- ifelse(danish$Sex == "M", -1, 1) * 0.5 
danish$cPrevError <- ifelse(danish$PrevError == "CORRECT", -0.5, 0.5)

lognormalParamMeanSigma <- custom_family(
  "lognormalParamMeanSigma", dpars = c("mu", "sigma"),
  links = c("identity", "log"), lb = c(0, 0),
  type = "real"
)

stan_funs_lognormalParamMeanSigma <- "
real lognormalmean2mu(real mean, real sigma) {
  real mu;
  if (mean < 25) {
    mu = log( mean + (exp(mean)-mean)/(exp(2*mean) + 1) ) - sigma^2/2;
  } else {
    mu = log( mean ) - sigma^2/2;
  }
  return mu;
}
real lognormalParamMeanSigma_lpdf(real y, real mean, real sigma) {
  return lognormal_lpdf(y | lognormalmean2mu(mean, sigma), sigma);
}
real lognormalParamMeanSigma_rng(real mean, real sigma) {
  return lognormal_rng(lognormalmean2mu(mean, sigma), sigma);
}
"
stanvars_lognormalParamMeanSigma <- stanvar(scode = stan_funs_lognormalParamMeanSigma, block = "functions")

m <- brm(RT ~ cSex + cPrevError + (cPrevError + 1|Subject), data = danish, family = lognormalParamMeanSigma, 
         stanvars = stanvars_lognormalParamMeanSigma, chains = 1)
summary(m)