I’m fitting the following with the
von mises as likelihood function:
form_ri_int <- bf(diff_theta ~ 0 + Intercept + emotion * mask + (1|id), kappa ~ 0 + Intercept + emotion * mask + (1|id)) prior_von_mises <- c( prior(normal(0, 2), class = "b", dpar = ""), # betas prior prior(normal(0, 1.5), class = "b", dpar = "kappa") # kappa prior ) fit_ri_int <- brm(form_ri_int, data = dat_fit, prior = prior_von_mises, family = von_mises(link = "tan_half", link_kappa = "log"), chains = 15, cores = 15, iter = 10000, file = "theta/fit_ri_int", seed = seed)
diff_theta is a variable measured in radians of which I’m predicting the circular mean and the spread (kappa).
I have some doubt about interpreting kappa given the link function. I’ve extracted my posterior draws for each parameters that I need and the doing
exp(kappa) otherwise some parameters can have negative values. Now the interpretation is that values around 0 means that the distribution is close to uniform and large values means less spread right?