I would like to see which factors can better explain the success of a particular event to happen. I'm interested in 4 factors, being before, during, season and observer. All 4 factors have only two possible outcomes (before: wet and dry; during:wet and dry, season: Winter and Fall) except observer, which has up to four.

I've organized my data the following way:

   success    before during    season observer
1:      no       wet    dry    Winter        1
2:     yes       wet    dry    Winter        2
3:     yes       wet    dry      Fall        1
4:     yes       wet    dry      Fall        4
5:      no       wet    dry    Winter        3
6:      no       wet    dry      Fall        1

My idea is to look at how this factors (predictor variables) best explain the response variable (success: yes or no) to happen. I would like to test this with and without interactions between predictor variables. All models will be constructed and the one yielding the lowest AICc will be the one best explaining a "success:yes" (e.g. success~before * after * season * observer).

Would that be the right approach given my data?

I'm unsure about what type of distribution to use, although I would assume binomial because only two possible outcomes are possible (yes and no).

Any other suggestions are welcomed. Also, is it necessary to include a random effect?

  • $\begingroup$ Logistic regression should do fine. What is the interpretation of the "observer" variable? $\endgroup$ Nov 20, 2019 at 17:22

1 Answer 1


Your can deal with your classification problem via logistic regression and compare your fitted models via their Classification Rate (or ROC Curve) on your test subset.

Regarding the response variable I would use a binomial distribution with n=1, so actually a Bernoulli distribution.


Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.