Hey all -
I am using Bayesian method multiple regression to assess life satisfaction. My predictor variables are all binary and have been converted to factors with 2 levels in order to assess the conditional effects. Below is an example of the code for my models:
Fit1 <-brms::brm(Life_Sat | weights(Weight) ~ V1 + V2, data = df, cores = 4)
Fit2 <- brms::brm(Life_Sat | weights(Weight) ~ V1 + V2 + V3 + V4, data = df, cores = 4)
Fit3 <- brms::brm(Life_Sat | weights(Weight) ~ V1 + V2 + V3 + V4 + V1xV3 + V1xV4 + V1xV2 + V2xV3 + V2xV4, data = df, cores = 4)
[note: x in place of and asterix which doesn't seem to appear on this forum]
All of the models came back with a number of pareto k values > .7 (fit 1: 11, fit 2: 15, fit 3:31) I reloo'd them and all pareto k values came back ok.
The issue and confusion I am experiencing comes from the order of best fit. Fit1 is suggested to be best, then Fit3 and lastly Fit2??
When checking for multicollinearity to try and make sense of what's going on I receive the following error message when attempting to check Fit3 (VIF and tolerance for Fit1 and Fit2 were fine)
Error in stats::vcov(x)[params, params, drop = FALSE] :
subscript out of bounds
Any insight about what could be going on or ideas of what I should do would be greatly appreciated!