The main goal of the present study was to discuss the application of the McNemar test to the comparison of proportions in dependent samples. Data were analyzed from studies conducted to verify the suitability of replacing a conventional method with a new one for identifying the presence of Salmonella. It is shown that, in most situations, the McNemar test does not provide all the elements required by the microbiologist to make a final decision and that appropriate functions of the proportions need to be considered. Sample sizes suitable to guarantee a test with a high power in the detection of significant differences regarding the problem studied are obtained by simulation. Examples of functions that are of great value to the microbiologist are presented.