McNemar test is a non-parametric test for a difference in proportion in two paired dichotomous samples.
The requirements of the test are:
Data in existing Excel worksheets can be used and should be arranged in a List dataset layout containing two nominal or ordinal scale variables. The observations of the two variables must be dichotomised by the same groups and criteria. If only a summary of the number of subjects for each combination of the dichotomous groups is available (contingency table) then a 2-way table dataset containing counts can be used.
When entering new data we recommend using New Dataset to create a new 2 variables (categorical) dataset or 2 x 2 contingency table ready for data entry.
To start the test:
Excel 97, 2000, 2002 & 2003:
Select any cell in the range containing the dataset to analyse, then click Analyse on the Analyse-it toolbar, click Compare Pairs then click McNemar.
The report shows the number of observations analysed, and, if applicable, how many missing values were excluded. The number of observations cross-classified by the two factors are shown as a contingency table.
The hypothesis test is shown. The p-value is the probability of rejecting the null hypothesis, that the samples have the same proportion in each group, when it is in fact true. A significant p-value implies that the samples have different proportions in each group.
METHOD An exact p-value is calculated using the Binomial distribution (see ). The 2-tailed p-value is computed as 2 x 1-tailed p
A point estimate and confidence interval for the difference of proportions between the samples can be useful to evaluate if the difference is practically useful. The point estimate can be expressed as a simple difference of proportions or as an odds-ratio.
To change the point-estimate and confidence interval:
The point estimate expressed as a simple difference or an odds-ratio and confidence interval are shown.
METHOD The confidence interval for the point estimate when expressed as a difference is computed using the exact Binomial distribution (see ). When expressed as an odds-ratio the logit Normal approximation is used (see ).