# Thread: Question- student paired sample t test, wilcoxon signed rank test

1. ## Question- student paired sample t test, wilcoxon signed rank test

Hi,

I have samples that are not expected to be normally distributed, I treat them with and without a compound and perform statistical analysis. I perform the Wilcoxon signed rank test that is for samples that are not normally distributed.
However, in some of my tests I do not have enough samples and I have therefore performed student paired sample t test instead. So my question is: If you perform student paired sample t test on samples that are not expected to be normally distributed are you then less likely to get a low p-value ( a significant different in the samples). So if you get a low p value(statistical significance), can you think that "oh, this is really significant since even the test that assumes normal distribution (which our samples are not) gives us a statistical significant p-value between the samples".

Anyone follow me on this? :-) Hope someone can comment and help me!

2. ## Re: Question- student paired sample t test, wilcoxon signed rank test

Tests

Ttest analog Wicoxon rank sum
paired ttest analog Wilcoxon sign rank

Your question is whether your run a ttest on skewed data, what affect it may have on the pvalue compared the pvalue for the analog test?

3. ## Re: Question- student paired sample t test, wilcoxon signed rank test

Thanks for your reply. Paired sample t test is not the same test as wilcoxon rank sum. Wilcoxon rank sum is for samples from different origin, where you cannot expect the same variations.
Paired sample t test is for normally distributed samles
wilcoxon signed rank test is for not normally distributed samples

samples from the same origin, treated with and without a compound.

I have only 3 and 5 samples

Originally Posted by hlsmith

Tests

Ttest analog Wicoxon rank sum
paired ttest analog Wilcoxon sign rank

Your question is whether your run a ttest on skewed data, what affect it may have on the pvalue compared the pvalue for the analog test?

4. ## Re: Question- student paired sample t test, wilcoxon signed rank test

hi,
maybe I misunderstand your situation, but the choice between a paired and a vanilla t-test is based on whether you have a way to build pairs for your before-after measurements. If you can not, or the pairs are arbitrary, then a paired t-test is not an option, simply because there are no pairs in your data.

regards

 Tweet

#### Posting Permissions

• You may not post new threads
• You may not post replies
• You may not post attachments
• You may not edit your posts