I’m using a logit model to evaluate the association of a set of habitats (H) to a particular land cover. Both land cover and H are binary variables.

glmer(land.cover ~ H1 + H2 + H3 + H4 + H5 + H6 + H8 + H14.1 + H14.2 + H14.3 + H15 + (1 |species)+(1|ISO_3), data=trainingData, family=binomial(link="logit"))

I want to know the probability of association of each habitat to the land cover. I'm not interested in the prevalence of the land cover. I know that the coefficients logit can be transformed into predicted probabilities using the following equation. However, this probability takes into consideration the prevalence of the land cover.

p= exp(β0 + β1*x1 + … + βk*xk)/(1+exp(β0 + β1*x1 + … + βk*xk))

I want to know which metric I can use to calculate the probability of association of each H ignoring the prevalence:

  • Can I use the coefficients of the logit model and ignore the intercept?

  • or Do I need to calculate the relative risk? how can I calculate the relative risk with logit coefficients and a multivariable model? Can I transform the relative risk to probability in a later step?

  • Are there other metrics that I could use to calculate this association?

In a later step I would like to transform the association of habitat to land cover to 1 and 0.



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.