Author(s): Bayan Shukr and Rhonda Magel

Abstract: Five tests for two proportions are proposed for use with data in a mixed design consisting of paired and independent data. The tests are combinations of the standardized versions of the McNemar test and the two-sample independent test for proportions using different weights. A simulation study is conducted comparing the estimated powers of the tests when the sample size for the paired data is equal to the sample sizes of the independent sample data (assuming equal independent sample sizes); when the sample size for the paired data is smaller than the sample sizes for the independent sample data; and when the sample sizes for the independent sample data are smaller than the sample size for the paired data. Correlations of 0.2, 0.3, and 0.4 are considered in the paired data. It is found that the test having higher powers when the paired sample size is larger is the test that assigns equal weights to both standardized versions or the test that assigns twice as much weight to the two-sample independent test. When the independent sample sizes are larger than the paired sample size, it is found the test having higher powers is the one that assigns more weight to the McNemar test. When the sample sizes are equal, the test that has higher powers assigns equal weights to both.

Pages: 105-109 | Views: 161 | Downloads: 3

Download Full Article: Click Here