# Interpretation of Interaction in Logit Model I am trying to interpret, for model 2, the betas for Success (Dummy 1/0). What is the probability females have success in technical fields and how do I calculate it with the betas.

m5 <- glm(success ~ female + technical + female*technical, data = subset(dt.data, company==0), family = binomial(link = "logit"))

m6 <- glm(success ~ female + technical + log(ex.goal) + backers + duration + currency + female*technical, data = subset(dt.data, company==0), family = binomial(link = "logit"))

• And how do I properly interpret Log likelihood and AIC in this model? – Phillip Forsgren May 31 '17 at 3:59

The linear predictor (alpha below) is computed like for a regression.

For model (1), this is

Male + non-technical: alpha = -0.729

Male + technical: alpha = -0.729 - 0.289

Female + non-technical: alpha = -0.729 + 0.208

Female + technical: alpha = -0.729 + 0.208 - 0.289 + 0.225 [interaction coef. comes into play because of the combination Female:technical]

Then the probability is: p = exp(alpha)/(1+exp(alpha))

For model (2) that depends also on the levels of the other variables (ex.goals, backers, duration). Just add their respective contributions to alpha like for a regression then compute their probability.

I hope this helps.