Good afternoon! I am currently working on my master thesis and I am a bit stumped on how to proceed practically with my analysis.
I am running a Multivariate Multiple analysis on a dataset, so I have multiple DVs and IVs, through R but after a few analyses I have confirmed that my dependent variables are highly skewed. Because of this, I had to switch from a linear model to a generalised linear model, specifically a Negative Binomial distribution. The problem is that now I don't know how to apply in R the MANOVA test statistics like Pillai's to glm objects.
An example of my code for the simple linear models was as follows:
model_1 <- lm(cbind(dependent_1, dependent_2) ~ Predictor_1 + Predictor_1*moderator, data = dataset_final)
summary(model_1)
Manova(model_1)
In this way, I had the significance/magnitude/direction of effects for each variable on each dependent, and then I used the car package's Manova() function to check if there were joint effects on the variables by checking if Pillai's test was significant. However, the glm.nb() function that I have to use to build the GLMs does not support cbind, so I have to create each model by itself in R, and so I cannot apply MANOVA to check for correlation in the coefficients. The code is now like this:
model_2 <- glm.nb(dependent_1 ~ Predictor_1 + Predictor_1*moderator, data = dataset_final, control=glm.control(maxit=50))
Summary(model_2)
model_3 <- glm.nb(dependent_2 ~ Predictor_1 + Predictor_1*moderator, data = dataset_final, control=glm.control(maxit=50))
Summary(model_3)
From what I understand, just applying the Anova() function to both models by themselves and check if a variable is significant in both won't help. I have looked at a previous similar question at here, but from what I can see there was no response on how to check for joint effect significance. How can I tell R to test if the variables in these different (as in, different objects) models have significant joint effects? Thanks a lot for your time.
