I am trying to setup a model in R. I want to test if strategies picked by different agents have a joint significant effect on the outcome.

My idea was to create a partial F-Test:

DV = success variable (0 or 1)
IV1 = agents
IV2 = strategies (around 10 types)
IV3 = agents_strategies (help variable created by me; a combination of each agent and each strategy they have chosen in the data)

L0 <- glm(DV ~ factor(`agents`), family = binomial)
L1 <- glm(DV ~ factor(`agents`) + factor(`agents_strategies`), family = binomial)
anova(L1, L0)

How can I account for multicollinearity? Can I set up my model in this way? Any help is greatly appreciated.

  • 2
    $\begingroup$ F-tests don't care about multicollinearity: you're good to go. $\endgroup$ Commented Jan 30, 2023 at 16:53


Your Answer

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

Browse other questions tagged or ask your own question.