What does it mean if my adjusted r-squared decreases between my bivariate and multivariate regression models? is this a bad or good thing?


1 Answer 1


The $R^2_{adj.}$ is a version of the $R^2$ which is adjusted for multiple regressors. Whereas the $R^2$ rises the more regressors you add, even when there are irrelevant, the $R^2_{adj.}$ stays robust.

$$R^2=1-\frac{SSR}{SST};\quad R^2_{adj.}=1-\frac{SSR/(n-p-1)}{SST/(n-1)}$$

Hence, when you add more regressors to your bivariate model, and then your $R^2_{adj.}$ decreases, the model tends to deteriorate and you might call it a "bad thing."

  • $\begingroup$ does this mean the bivariate model is better? Z is the third regressor or a confounding variable. $\endgroup$
    – Louis Rich
    Commented Dec 17, 2020 at 19:00
  • 1
    $\begingroup$ @LouisRich Yes, I would consider that model with the higher $R^2_{adj.}$ better, because it explains more variance in the dependent variable. Or put the other way round, the added regressors contribute nothing but "blur" to the model. $\endgroup$
    – jay.sf
    Commented Dec 17, 2020 at 19:05
  • $\begingroup$ that's perfect, thank you so much. So would this mean that Z is not a confounding variable? and would provide support for hypothesis in the bivariate model? $\endgroup$
    – Louis Rich
    Commented Dec 17, 2020 at 19:07
  • $\begingroup$ @LouisRich I would'nt conclude that solely from the $R^2_{adj.}$. The effect of the $X$ variable is probably more important (does it change considerable? does it change sign? does it get insignificant?). $\endgroup$
    – jay.sf
    Commented Dec 17, 2020 at 19:21
  • $\begingroup$ HI X in the bivraite model 1.2052, STD error 0.1603, p value= 2.58e-12 ***. $\endgroup$
    – Louis Rich
    Commented Dec 17, 2020 at 19:25

Your Answer

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

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