1

I have built a binomial regression model using the equation:

logit_1 <- glm(past_adl_binary~., family = binomial(link = "cloglog"), 
               data = newdata)

The outcome of the model is as follows: enter image description here

Can anyone help me interpret the outcome of this model? I understand that all of the coefficients are significant apart from sunangle.

Can I use this equation (100*exp(parameter estimate)-1) to calculate the increase/ decrease in percent probability of my y-variable (1 or 0) changing based on my predictor variables? For example, for habitat_categoryONDEEP: 100*(exp(-0.173)-1) = -15.886. Would this suggest that a habitat category of ONDEEP decreases the probability of my y-variable being 1 by 15.9%, compared to the reference category?

Sycorax
  • 90,934
Beth Edwards
  • 11
  • The size of the AIC is irrelevant: all that matters are differences between the AICs of related models. In fact, different software platforms may report different values of AIC for a given model and dataset! – whuber Mar 28 '21 at 16:51
  • The cloglog link does not follow the rule of thumb that you state above; it works on the log-hazard scale: https://stats.stackexchange.com/questions/132627/interpreting-estimates-of-cloglog-logistic-regression (this might get closed as a duplicate of that answer) – Ben Bolker Mar 28 '21 at 23:23

0 Answers0