Hello,
I am new to bayesian methods and currently running the following logistic model on 10 imputed datasets.
All explanatory variables are factor variables. region has 8 levels and period has 3 levels. After running the model, I am interesting in checking if the response for the second and third period is different compared to the reference first period for each region.
I have run several hypothesis tests (similar to one shown below). Exponentiating the estimate and CI intervals from the hypothesis test results provides odds ratios consistent with those from frequentist analysis.
mod <- brm_multiple(
case ~ region * period + imd + age_group * vaccinated + lab_result,
data = data,
prior = set_prior("normal(0, 1)"),
sample_prior = TRUE,
family = bernoulli()
)
hypothesis(mod, c(region_2 = "period3 + region2:period3 = 0"))
Hypothesis Tests for class b:
Hypothesis Estimate Est.Error CI.Lower CI.Upper Evid.Ratio Post.Prob Star
1 region_2 0.26 0.13 0.01 0.53 1.51 0.6 *
---
My question is 1) whether this approach is appropriate and 2) are there any other approaches recommended instead of the hypothesis function? I have read somewhere the posterior draws can be used for such analysis but I am unclear how I would be able to apply the contrasts correctly on posterior samples and obtain odds ratios?
Thanks in advance for any advice.