Health Sciences 3801A/B Lecture 7: Lecture 7 - Dependent Groups t-Test
Document Summary
If the correlation is 0, use the independent groups t-test. If there is a correlation that is positive, the t-test is larger because you are pulling out variance that is accounted for by some confound within the design. If you match for another characteristic, you are allowing the removal of the confound within the design. The denominator gets smaller because the error is getting smaller. The n-sizes are equal because the values are sampled in pairs. The df = n (number of pairs) - 1. The formula is exactly the same as the single sample t-test.