# Thread: Binary predictor in regression analysis?

1. ## Binary predictor in regression analysis?

I have just had a lecture with a lecturer summarizing regression analysis, and showed us some examples. One of those examples included a regression with interaction including a binary predictor, e.g.

y = a + bx + cx + bcx

Where b was either 1 or 0. I thought that was pretty weird, but cannot recall any specific reference on deeming it fine or wrong. I also asked our lecturer if he could tell what would happen with the predictor if b was a continous variable like the rest (and not binary), and he simply answered that he did not know.

What do you say?

2. ## Re: Binary predictor in regression analysis?

Many times you will see binary predictors in interaction terms.

Say predicting weight and you had height and gender. This would mean that weights regressed on height would be different for the two gender groups (different slopes, males heavier in health individuals).

Now say you were predicting the days sick a person was and you had number of cigarettes smoked and particulate matter in the air. there is a chance that individuals who smoke have more sick days as PM increases beyond the individual effects of either alone.

3. ## Re: Binary predictor in regression analysis?

I don't understand your lecturer's comment. Nothing would change at all if b was continuous. It might be more difficult to map out the interaction effect graphically, but that would be all. Dummy variables as predictors are identical to interval variables (you interpret the slopes somewhat differently, but that is all).

4. ## Re: Binary predictor in regression analysis?

See the first picture in post #4:

5. ## Re: Binary predictor in regression analysis?

Aha. It just seems so out of place. If the regression included genders, why not just make two regressions? Wouldn't it be easier and more managable?

Originally Posted by hlsmith
Say predicting weight and you had height and gender. This would mean that weights regressed on height would be different for the two gender groups (different slopes, males heavier in health individuals).
Is this the same as examining parameter heterogeneity/homogeneity?

 Tweet

#### Posting Permissions

• You may not post new threads
• You may not post replies
• You may not post attachments
• You may not edit your posts