1. Each subject is asked several yes/no questions, and the only data available are the proportions responding yes. If the cell counts of the entire 2x2x...x2 table are available, you can use Cochran's Q. See the FASTats entry on Cochran's Q.

2. Each subject is categorized into one level of a multilevel variable, and you want to compare the proportions in each level.

Berry and Hurtado (1994) presented SAS/IML programs that give confidence intervals for the difference in dependent proportions. Use the MULTIBIN SAS/IML module for scenario 1. See this discussion of scenario 2.

_________

Berry, J.J. and Hurtado, G.I. (1994), "Comparing Non-Independent Proportions,"[POSTSCRIPT] Observations: The Technical Journal for SAS Software Users, 3(4), 21-27.

Operating System and Release Information