in my ANCOVA analysis, I have the case that my Y and X show a nonlinear pattern. Transforming either only Y or X does not help to stabilize variance. I did not find any hint that doing an ANCOVA in log-log space is illegal, however, I also did not find any note on that one should try out a transformation of Y AND X...

Could someone give me a hint if I am doing wrong with log-log?


  • 1
    $\begingroup$ Because ANCOVA is just a form of linear regression, you will find useful replies by searching on the regression and data-transformation tags. $\endgroup$ – whuber Apr 11 '12 at 15:34
  • 2
    $\begingroup$ Thanks, I found out that log(Y+2) already helped to stabilize the residuals. I used the logtrans function from the MASS package in R to come to this conclusion. $\endgroup$ – Jens Apr 12 '12 at 16:30

I found that sometimes in necessary to "linearized" the relationship between X and Y. if you have an allometric relationship or for instance a power function, a log-log trasnformation help to do that and also ofcourse help to stabilize variances. it all depends in the nature of your data

| cite | improve this answer | |

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.