2
$\begingroup$

I'm working on a question on 2-way ANOVA where there are 2 treatments and 1 blocking factor. The residual plots suggest unequal variance. What is another parametric or non-parametric alternative method to 2-way ANOVA?

$\endgroup$
2
  • $\begingroup$ If you have one blocking factor and one other factor, then Friedman test is the non-parametric alternative to classical 2-way ANOVA with blocks. How did you assess variance heterogeneity? Usually, variance tests are conducted after having accounted for block effect (i.e., tests are performed on raw data adjusted for block means). $\endgroup$ – chl Dec 13 '12 at 22:34
  • $\begingroup$ The generalized p-value approach: www3.stat.sinica.edu.tw/statistica/oldpdf/a7n35.pdf $\endgroup$ – Stéphane Laurent Apr 13 '13 at 7:44
2
$\begingroup$

ANOVA is equivalent to regression using categorical (aka "dummy" or "indicator") variables. So any techniques from regression can also be applied to ANOVA. Here are some suggestions:

  • If the error terms appear to be normally distributed, you could try using weighted least squares.
  • If the error terms are not normally distributed, you would probably need to transform the variables.
$\endgroup$

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.