Can I include all control variables to the hierarchical model in brms? Does Including several variables at the same model making it overfitting?

Hello All,

I hope my message find you well

I benefited a lot from your valuable answers Prof. @paul.buerkner. I needed to write new topic not to mix topics together. I also needed to explain the context related to the code by presenting a brief on my work

I am using the brms package to conduct a Bayesian multivariate multilevel model for my PhD thesis, a cross-national study in 7 countries. My aim is to discover how subjective well-being can affect the modes of unconventional political participation. In addition, I use the country’s score of voice and accountability as moderator.

IVs : Life satisfaction, financial satisfaction, health, happiness, free choice.

DVs: signing petition, joining peaceful demonstration, joining strike, joining boycott.

Control Vs: Employed, Retired, Student, Unemployed, Male, careful_trust, age.

Country-level Variable: country’s score of voice and accountability.

Number of Observations is 13000.

Priors: non-informative priors, using the default settings.

May I ask several questions about the model fitting?

Here is my code:


fit1 ← brm(
mvbind(sign_pet, join_boycott, attend_demo, join_strike) ~ satisfaction + financial * var + health * var + happiness * var + choice * var +
Employed + Retired + Student + Unemployed +
Male + careful_trust + age + (1 |p| country),
data = final_dataset1_scaled,
family = cumulative(link = “logit”),
chains = 4,
cores = 4,
iter = 2000,
control = list(adapt_delta = 0.99, max_treedepth = 15)
)
summary(fit_life_satis)


Q1: Does big number of variables necesserily make overfitting issue? or is it not causally related?

Q2: Is it a must to include all the control variables at the same model? or can I fit separate models for each variables?

Q3: Is it correct to include the control variables similar to the IVs before (1 |p| country) this way:

~ satisfaction + financial + health + happiness + choice +
Employed + Retired + Student + Unemployed +
Male + careful_trust + age + (1 |p| country)

or

Do I have to assign them after (1 |p| country), this way:

~ satisfaction + financial + health + happiness + choice +
Employed + (1 |p| country) + Retired + Student + Unemployed +
Male + careful_trust + age ?

Q4: I am including interaction with group-level variables “var” (countries’ score of voice and accountability), does repeating the interaction with each IV is right?

Your answers will be highly priceless.

Respectfully,
MD