#### Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!

Supported by

# Bayesian multiple Regression - compare regression coefficients

Hi everyone,

I would like to test whether the size of my regression coefficient varies across three different groups (A, B, and C). The goal is to find out whether the regression coefficient of weight predicting insulin levels differs across three groups of different age.

Therefore I computed two dummy variables: V_A: coding 1 for group A and zero otherwise, and variable V_B: coding 1 for group B and zero otherwise. Now in order to test whether the regression coefficients vary across groups I computed the variables: V_A_weight = V_A*weight and V_B_weight = V_B*weight.

In the Bayesian regression model I enter "insulin levels" as dependent variable, and "weight, V_A, V_B, V_A_weight and V_B_weight" as predictors. Because my main interest is to see whether a model including "V_A_weight and V_B_weight" is more plausible I assigned "weight, V_A, V_B" to the null model.

Is this procedure correct?

is it correct to only focus on the model including "V_A_weight and V_B_weight" as additional predictors?

do I need to include V_A and V_B at all into the regression model?

Any help would be very much appreciated!

• Quick response: my first thought it to include both weight and group as predictors, and then add the interaction (which is of interest)

Cheers,

E.J.

• Thank you for the fast response!

But group is a nominal variable and therefore I can not enter group as a predictor, right? That's why I created the other variables.

Thank you!

• You could call the groups 1, 2, and 3 and pretend you're OK (this will violate assumptions, but when you are wiling to add VA and VB as predictors this should not worry you perhaps)