Nominal vs Nominal
Introduction
If you have two nominal variables, each with the same categories (for example with a before - and - after type of experiment) you might want to know if the responses have changed or there is a shift in opinion.
Let's say you asked people which brand they prefer out of five (labelled A till E), then show them a commercial and ask the same question again. You can now either test if the percentages for each of the five brands have changed overall (e.g. 20% said brand A before the commercial, but 60% said brand A after). To test this you can use a Stuart-Maxwell test or a Bhapkar test.
If you want to know if a shift took place, for example 10% shifted from A to B, but 30% shifted from B to A, then you can use a McNemar-Bowker test.
For either of the above you are here at the right section. If however you are simply interested if two nominal variables have a relationship between them (if one might have an effect on the other), have a look at the unpaired version of two nominal variables here.
The analysis is split into a few different steps:
1) Get an impression of the sample data by creating a cross table
2) Visualise the sample data in a clustered bar-chart
3) Perform a test (Bhapkar or Bowker) to say something about the population.
4) Write up the findings.
For each step a different page was created, so let's begin with step 1 on the next page.
Google adds