# Step-wise Multiple Regression or ANCOVA

I have an assignment that gives a dataset and a choice of 3 tests: Step-wise Multiple Regression, ANCOVA and Log-Linear Analysis.

The dataset consists of Blood Pressure measurements as the dependent variable (DV), and Smoking (True/False) as the independent variable (IV) with confounders being Age and Gender. My hypothesis is that smoking increases blood pressure but being a man or being older would make blood pressure increase too.

I have ruled out Log-Linear Analysis, but I am getting confused which of the other tests would be more appropriate. If it was just Age as a confounder I would do ANCOVA as it is a continuous covariate but I'm not sure what to do with Gender as it's categorical. As I understand it, ANCOVA and Multiple Regression are identical mathematically so the distinction and implementation is making me completely lost!

Would greatly appreciate if anyone could help me clear this up? Thank you

• As an aside, I'll never understand why teachers assign questions like this. It's never really important to be able to distinguish between a "multiple regression" model or "ANCOVA" model in the real word, for example. What's important is that you are able to state your model and ensure that the data you've collected are appropriate for your model and it's assumptions. Too much time is wasted focusing on this type of minutiae. – StatsStudent Dec 20 '18 at 21:08
• To help you answer this question, consider if one of your model choices is a subset of the other. – StatsStudent Dec 20 '18 at 21:13
• Couldn't agree more, to put it in context it's a Masters level stats class that has not touched regression or the concept of modelling at all. Thanks for the edit! – Oliverw31 Dec 20 '18 at 21:14
• Hopefully my second comment helped you out a bit more. Is it clear now? The first one was just me venting ;-). – StatsStudent Dec 20 '18 at 21:20
• I think ANCOVA is a subset of multiple regression and so if my model uses categorical predictors like gender and continuous predictors like age, I can just call it multiple regression and do it in a step-wise way to make my lecturer happy? Am I on the right lines? – Oliverw31 Dec 20 '18 at 21:21