Re: paired observations

In all cases I know the paired t test (on a computer) will simply drop
the cases with missing data. I'd be interested in how you'd run the
paired t test any other way (though I can think of ways it is possible
.... e.g., by imputation of missing data). Whether it is appropriate
depends on the reasoning for missingness ... you may need to look up
terms like MAR (missing at random) or MCAR (missing completely at
random).

Thom

.

Relevant Pages