Regression Episode 7: Interactions of Categorical with Continuous Predictors

In previous episodes, the effects of all predictors have been assumed to be additive. In this episode, Dan considers interaction effects, focusing on interactions between categorical and continuous predictors...

Dan describes how interactions are captured through the inclusion of product terms between the continuous predictor and the coding variables for the categorical predictor. He describes how the effect of the categorical predictor then depends on the value of the continuous predictor and vice versa. The effect of one predictor at any given level of the other, called a simple slope, is often depicted by a conditional effects plot. Dan describes how this plot is generated and used to understand and test interactions via an example looking at the prediction of course quality ratings from the easiness of the class and whether the professor is perceived to be attractive or not.

[larissacury]

Hello, again! Do you have any recomendations on categorical * continouns interactions but this time with a 3-level categorical predictor as you've explained in the video w/ the 2-level one? I know how to dummy code the categorical main effect betas, but i'm struggling what understanding how the interaction betas will be

[larissacury]

Hi! I'm having a hard time: I have a regression with Y ~ YEAR * TEST(A/B). It turns out that YearB has a significant slope on the ref level (Year 1, Test A), but Test (B) doesn't. However, the interaction Year * Test is still significant (Year2, LangB, negative significant interaction), any headings on this?

[larissacury]

I'm here once more, thank you very much! The video is amazing! Let me ask something that I haven't thought before: what's the difference between performing a Niemann test and looking at the confident intervals of B3? Also, would you give a hint on how to report these results?

[larissacury]

Would you give us some hint on how to report this results? I'm currently having some trouble. I have posted a question yesterday on CrossValidaded called "How do we report a regression with categorical * continouns interaction?" (Youtube won't let me post the link for it). If you may, would you take a look on that? I'm just wondering how to report the interaction (b3) and b2 effects

[larissacury]

Hi! What if we have NA/missing values in the continuous predictor (which is interaction with the categorical) and in the outcome continuous variable?

[larissacury]

Thank you! What happens if the interaction is not significant, should we consider it or not?

[FelipeDias-wobw]

Great explanation. The trick is when the B3 coefficient is not significant, but the conditional effects are. Any suggestions on how to make sense of it?