One of my hypotheses is that Variable A and B predict Variable C, so I want to perfom regression analyses. My questions for this example:

1) Should I perform two separate simple linear regressions or a multiple regression with both A and B as predictors? What is the difference? (A and B do not correlate with each other).

2) Scatterplots for the associations of Variable C with A and B respectively look pretty random.

Correlations accordingly are also not significant. If I already now this, does it still make sense to do regression analyses? Besides linear regression, would another type of regression model be better to use? What should I report in my paper in that case?


A couple of thoughts:

  1. It is usual to put the dependent (or outcome) variable on the vertical axis and the independent (or predictor) variable on the horizontal axis

  2. If you do two separate simple linear regressions, you are, essentially, fitting a line to the plots you have posted. I agree that it doesn't look like there is much there. If you do a single multiple linear regression, then you are looking at the relationship between A and C, controlling for B and between B and C, controlling for A. That may show something that you didn't find here.

  3. Something is wrong with your plots, as C has very different values in the two plots 0 in the one on the left, it ranges from 0 to 15, in the one on the right, it ranges from 200 to 320.

| cite | improve this answer | |
  • 1
    $\begingroup$ Thank you for your comment. While editing the scatterplot to post it here, I labeled the axis wrong, so thank you for pointing that out. I edited the post and now the labeling is correct. $\endgroup$ – mcfilu Oct 19 '19 at 16:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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