Generalized Linear Model: help interpeting interaction


I apologise in advance if I missed it, but I couldn't find an answer in previous threads.

I run a generalized linear model with 1 factor (gender) and 5 covariates. No single variable appears to have a predictive value on my VD. However, when I test for interactions, many seem to have a significant effect.

First thing I thought is that probably a single variable has not enough predictive power over the VD, but it does when is combined with other(s).

Does this make sense? Theoretically, I can find an explanation to this, but how should I report it? «Individually, variables had no predictive effect on the VD but when considering the combined effect, they did»?

Thank you so much in advance =)


Less is more. Stay pure. Stay poor.
Can you really predict Gender?

Well there could be a synergistic interaction on the multiplicative scale. So one group in the presence of another results in a greater likelihood of the outcome. This is completely feasible, though we don't know your context so who knows if it makes sense. What you should do is calculate the probabilities for the groups along with the odds ratios with 95% CI. This way you can see the effect.

P.S., When doing this, did you leave the main terms in the model, so y = x1 + x2 + x1*x2 ? Which is highly recommended.