How can I compare the means of two data sets where there is some overlap in the samples so that the groups are neither paired nor independent? That is there has been some change in the samples with some subjects providing data in BOTH sets, some respondents only providing data in the first set and others only providing data in the second set?

  • $\begingroup$ Can you give more details about the practical situation giving rise to this? $\endgroup$ – kjetil b halvorsen May 17 '16 at 13:54