Matched pairs vs Each pair Student's T

[#1]

Hello, 

I am trying to look at the effect of two conditions on my matched paired data. I am confused as to what the pair Student's T test does (Analyze > Fit Y by X > Compare Means > Each Pair, Student's T) vs Analyze > Specialized Modeling > Matched Pairs

I am trying to understand what the Each Pair, Student's T does and when I should be using that vs Matched Pairs.

<JSL> 15.2.1

[#2 · Accepted solution]

The compare means, each pair, Student's T is equivalent to the 2 sample pooled t test.  It treats the groups as independent groups, but pools the data to estimate the standard error of the difference between the groups.  The Matched Pairs test is a one-sample test, so if you really have matched pairs that is the correct platform to use.  However, I would note that I am not a big fan of matched pairs - there are few natural settings and when the matching is accomplished by constructing 2 groups that appear to be matched, I'm skeptical about using the matched pairs approach.  When it is data about the same individuals, say before and after a treatment, then matched pairs can be appropriate (provided that little else has changed other than the treatment).  But the Compare Each Pair assumes that there is no ordering in the data, it is just random samples from different groups.

[#3]

Thanks Dale, that makes so much more sense now