1

This must be common knowledge and asked so many times, but I can't find it. I'm a bit of a rookie, so my apologies if it's a redundant question. I am doing a linear regression with a binomial independent variable and a continuous dependent variable, and can't figure out how to test my model fit. QQ plot is fine, other than that I'm not getting anywhere.

I saw this: How to perform residual analysis for binary/dichotomous independent predictors in linear regression? but do not understand if it's applicable for my bivariate and how to try.

Could someone please help me?

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
c.lucette
  • 11
  • 4
  • Can you clarify what you mean by test your model fit here? Are you asking for a test of the model as a whole such as a global F-test for the model? Also, regression models don't make assumptions about the distribution of the independent variable, so it isn't clear there is really an issue here, but what exactly do you mean by "binomial"? Do you simply mean *binary*, ie, is this a t-test? – gung - Reinstate Monica Feb 26 '16 at 23:07
  • My apologies. I just want to test if a linear regression is an appropriate model, or if I should transform, add a variable, etc. Usually, I would mainly look at the residuals, but that is not possible. I mean that I have a dichotomous variable, which is indeed binary. I am just struggling with checking if a linear regression works for my data. – c.lucette Feb 26 '16 at 23:35
  • If you have a linear regression w/ a continuous Y-variable & a dichotomous X / independent variable, then you have a t-test. You can examine the residuals to assess the assumptions of your model. – gung - Reinstate Monica Feb 27 '16 at 00:10
  • Yes, I figured that's a t-test but for the assignment we cannot directly use a t-test. My question is, how do I examine the residuals? Because obviously I just have two vertical lines of residuals on my plot, and I do not know how to deal with that. Thanks! – c.lucette Feb 27 '16 at 00:21
  • Please add the `[self-study]` tag & read its [wiki](http://stats.stackexchange.com/tags/self-study/info). Then tell us what you understand thus far, what you've tried & where you're stuck. We'll provide hints to help you get unstuck. – gung - Reinstate Monica Feb 27 '16 at 00:24
  • Added the tag, read the wiki. Thank you! I thought that I had to check for linearity, residuals, etc. My assignment actually asks for a comparison of residuals. I have tried to plot my model, and check residuals, but I do not know how to interpret them and if I even can interpret them. I have read as an answer on the linked question to do "a regression on a single binary predictor and fire up quantile box plots comparing the residuals for the two levels of the predictor". I do not understand what that means or how to do it. I have done QQ plots, but I don't feel like it's sufficient. – c.lucette Feb 27 '16 at 00:41
  • The qq-plot checks the normality of the residuals. It doesn't mean anything to check for linearity if you have only 2 points on X (ie, it's binary). That just leaves the question of constant variance; you can check the residuals for that just as you would for a linear regression w/ a continuous IV. – gung - Reinstate Monica Feb 27 '16 at 00:48
  • So I would look at the vertical spread to check constant variance and compare the vertical spread for x=0 and x=1? Thank you so much for your help, I really appreciate it. – c.lucette Feb 27 '16 at 09:56

0 Answers0