I need to respond to a reviewer's critique of my use of an adjusted Wald F Test (extension of McNemar test) to examine change in within-group proportions over time using weighted analyses (svy commands). The situation is this: We have a dichotomous health measure that we administered at two timepoints. Our goal was to compare the proportion who endorsed the condition at T1 to the proportion who endorsed the condition at T2 while applying survey weights (svy commands) and recognizing the within-group nature of the data (same people responded at both timepoints). Based on our reading of the literature, we elected to run a Wald F Test (extension of McNemar test), which compares proportions for the discordant pairs (i.e., it compares the % who who endorsed yes at T1 and no at T2 to the % who who endorsed no at T1 and yes at T2). A reviewer has said that this was the wrong approach because there could be a large discordance between endorsement at the two timepoints (i.e., there could be many people who endorsed yes at only one timepoint), which would erroneously result in a high Wald F-test. However, I'm not sure that this is actually the case. The reviewer gave the following example: Consider the hypothetical example, of a two-by-two table matching the % in Physical Health Status (at least one Physical Health Condition=Yes): T1=Yes/T2=Yes: n=78, T1=Yes/T2=No: n=454, T1=No/T2=No: n=0, T1=No/T2=Yes: n=468. The proportion at T1=Yes is 53.2% and T2=Yes is 54.6% which are not far apart, however, there is a large discordant (922 out of 1000), which will result in high Walt F-test (rejecting the null hypothesis). Can anyone provide any input on whether this reviewer is correct or not, and if they are correct, can you suggest another analytical strategy for this scenario? We originally planned to apply the McNemar test but couldn't figure out how to do that appropriately with the svy commands. Thanks in advance for any help!