Chi test and regression

[Englund]

Yes Dason, of course. Why not a fan of transforming data? If you have small samples and wish to test a hypothesis using a t-test, don't you think then a good idea is to transform the data and hope for a normal distribution? If anything, that thing could serve as a 'back up'-test to your other tests that u perform.

But I agree, you should not perform tests on transformed data unless it is really necessary.

[Dason]

If you have small samples and wish to test a hypothesis using a t-test, don't you think then a good idea is to transform the data and hope for a normal distribution?

No.

Don't get me wrong. I'm fine with transformations if you have a reason for them or they're appropriate for the type of data you're working with. But transforming just to achieve normality is silly to me. If you're doing that then why not just directly the model the data you have with an appropriate distribution? But transforming and then doing some sort of test is not a good route to go in my opinion. Typically the interpretation of the parameters of interest aren't the same once you transform the data. Heck sometimes its hard for me to understand what the parameters you're estimating actually represent once you transform the data. I don't mean to sound pompous but if I have trouble understanding the meaning of the parameters once you transform the data then I sure don't expect most people to understand what the parameters actually represent.

There are all sorts of tests you can do. Why force your data to meet a normality assumption when all you're really doing is making it sort of look kind of normal maybe and then applying a test appropriate for normally distributed data. If you don't think the data is normal then do a nonparametric test. Or do a randomization test. Or model the data with an appropriate distribution. But it seems like "it's the only thing I know how to do" is the only actual reason people resort to transforming the data and doing a t-test.

If anything, that thing could serve as a 'back up'-test to your other tests that u perform.

But I agree, you should not perform tests on tranformed data unless it is really necessary.

Why are we doing "back up" tests? This sounds like you're saying that if we can't find signficance with the test we thought was actually appropriate for the data then we should just try this other test and maybe we'll find significance? That sounds a little like fishing to me. I'm not a fan of fishing.

[Englund]

Why force your data to meet a normality assumption when all you're really doing is making it sort of look kind of normal maybe and then applying a test appropriate for normally distributed data. If you don't think the data is normal then do a nonparametric test.

There are situations when non-parametric tests wont work well. I had a situation like that some time ago. I had two small samples where I wished to test for equal sample means between the samples (three groups in each sample). In this case non-parametric tests would suffer from low power like hell, and they would be incorrect to use because of too small groups (if we would have done Chi square test for example). In this case I think it was necessary to transform some data. Fortunately, I only had to transform one column of data for it to look like a multivariate normal distribution.

[Karabiner]

In case of very small sample size, parametric tests suffer
from low power either. And in case of very small samples
it is difficult or impossible to support the claim that
the transformed data stem from normally distributed
populations. The tests for normality are underpowered
and the graphical displays of the data often are
inconclusive. Similar problems may arise with other
assumptions of a parametric test.

With kind regards

K.