I am trying to fit a model to predict a quantitative response variable, using several binary variables. In particular, I am interested in measuring the relationship between one of the binary variables and the outcome variable, holding the other binary variables constant.

My inclination is that ANOVA is the appropriate method to use. The advantage of ANOVA over a series of t-tests should be avoiding Type 1 errors. But I seem to recall from stats that ANOVA is primarily designed for predictor variables with more than 2 levels, aka not binary.

Which is preferable?

Final question: is there any reason not to use multiple regression in this case? Thanks.

  • $\begingroup$ @NickStauner Okay, moved to an answer at your suggestion, and slightly edited. $\endgroup$
    – Glen_b
    Jul 24, 2014 at 11:19

1 Answer 1


ANOVA is fine with binary predictors.

Binary predictors are common, even.

Multiple regression will be equivalent - you can fit the same models (including interactions, if needed) either way. Whatever suits you.

If your quantitative response is a count, you might consider a Poisson or binomial GLM, and if it's continuous but right skew you might consider (say) a gamma GLM.

(See the answers here for example.)


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.