Ordinal vs Ordinal paired
If you have two ordinal variables that use the same categories you might either be interested in knowing if there is a relation between the two variables (i.e. if one increases does the other increase or decrease), or if there are changes (differences) between the two (e.g. did many people agree with the first, but disagreed with the other). To test for a relation you can use the technique used at unpaired discussed on this page, however if you are interested in differences this section is the place to be.
To test for a difference we will use the following steps:
1) Start with getting an impression from the sample data by creating a cross table and determining the median for each variable.
2) Visualise the sample data
3a) Test if there are any differences perhaps also in the population using a Wilcoxon signed rank test and
3b) determine the magnitude of the difference using an effect size measure.
4) Combine the results to create a report.
For each of the above steps a separate page is created. Let's begin with the impression of the sample data on the next page.