non-parametric equivalent to 2x2 mixed

#1
Thanks to anyone who reads this post.

I'm not very familiar with non-parametric tests, however after looking at my data and running Levene's test it is apparent that I have to use some non-parametric tests. I'll give a little background on my research first. Basically I have about 250 participants which I've categorized into two groups Sensitives and Non-Sensitives. Sensitives=66, NonSen=184. Both groups visited 8 rooms however the rooms where categorized into High and Low based on the amount of electrical fields. The participants were asked to report if they expereinced any physical effects in each room. We were looking at wheter the participants would have more physical compliants in the High EMF rooms.

So basically each participant visited 8 rooms and marked yes or no for each room, with the rooms either being low or high and the participants either being sensitive or non.

...........................High Room..............Low Room

Sensitive....................#......................... #
Non-Sensitive ............#.......................... #


1. Is there a nonparametric equivalent to a 2x2 mixed designs anvoa?
2. If not any suggestions on how to compute the data?
3. What are the non-parametric tests that are equivalent to a 2 way between subjects Anova and a two-way within subjects Anvoa?

Thanks for any help
 

Dragan

Super Moderator
#2
Thanks to anyone who reads this post.

I'm not very familiar with non-parametric tests, however after looking at my data and running Levene's test it is apparent that I have to use some non-parametric tests.
Thanks for any help

I think it might be better if you considered some kind of variance stabilization transformation. Or, use a Welch (type) adjustment to the degrees of freedom.

That said, because you have a 2 x 2 design, you can take the ranks of the original data points and perform the usual OLS parametric proceedure on ranks. This is what is known as a rank transformation. However, this is only legitimate for 2 x 2 designs.