I'm stuck on a question regarding t procedure. I must determine if the plot (Q-Q normal) is suitable for t procedure. Second question asks which procedure, the pooled-variance t procedure or Welch t procedure, should I use.

I have two sets of population, independent of each other. There are no outliers, n=20.

Question

I'm stuck on a question regarding t procedure. I must determine if the plot (Q-Q normal) is suitable for t procedure. Second question asks which procedure, the pooled-variance t procedure or Welch t procedure, should I use. 

I have two sets of population, independent of each other. There are no outliers, n=20.

anonymous · Answer

Welch presumes unequal variance while pooled-variance assumes the populations have the same variance

anonymous · Answer

the Q-Q plot is just to ensure our distributions appear 'nice' and approximately normal (few outliers, nice linear trend, etc.) -- it's for justify the statistical methods you use

anonymous · Answer

so that's how you decide whether to use Welch or pooled variance? I thought there were some kind of test or methods to decide. What to do when variance is unknown?