StatTools : Sample Size for McNemar Test of Paired Changes in Proportions
Explanations and Tables

Links : Home Index (Subjects) Contact StatTools

Related link :
McNemar Test for Paired Changes in Proportions Explained Page
Sample Size for McNemar Test of Paired Changes in Proportions Program Page

Introduction Sample Size Tables References
The tables on this page present sample size (in number of paired observations) required to detect differences between proportions of changes in directions. The parameters are
  • Power (1-β), where β is the Probability of Type II Error, so that Power is the probability of detecting a significant difference if it truly exists. Power values of 0.8, 0.9, and 0.95 are provided, each in a separate table. The most common value used for Power is 0.8
  • α, the probability of Type I Error to be used to decide statistical significance. α of 0.1, 0.05, and 0.01 are provided in the tables, although the most common value used is α=0.05
  • p+- is the probability or expected proportion of a change from Positive (+) to Negative (-), and p-+ is that in the other direction. Positive and Negative are merely terms used to indicate opposing directions, and the effect is symmetrical for calculating sample size. For examples the sample size for p+- = 0.2 and p-+ = 0.4 is the same as p-+ = 0.2 and p+- = 0.4
  • T_1 and T_2 in these tables represent the One Tail or Two Tail models. In most cases, the Two Tail model is used in the McNemar Test, as most researchers are interested in whether a significant difference in directions of change occur without worrying about which direction. However, should a pre-determined direction is of interest, the one tail model can be used as it is more powerful (requiring smaller sample size).