I use R. I have a linear mixed model and I use its output to estimate my posterior distribution (with uninformative priors) using the function sim.

I calculate the effect size between different subgroup as differences between the fitted values. Shall I use the same criteria also for the credible intervals (CrI)?

Example: Subgroup1 has (-0.11, 0.01, 0.13) as lower, fit and upper estimates (95% CrI) Subgroup2 has (0.09, 0.26, 0.42) as lower, fit and upper estimates (95% CrI)

The effect size is for me 0.25, what about its CrI? Again just the difference?


1 Answer 1


Rather than subtracting 0.01 from 0.26, you should be using all of the posterior estimates from which you can obtain the posterior effect size estimate (mean/median) and interval. For instance, generate thousands of estimates for each coefficient, generate a new variable with the difference between the entire posterior distribution of two coefficients, and then use the quantile function to obtain an equal tailed credible interval.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.