topic Re: Matched pair analysis with categorical responses in Discussions

Matched pair analysis with categorical responses

GregMcMahon — Fri, 09 Jun 2023 00:23:32 GMT

Hello everyone,

This will be straightforward I'm sure. I have 100 rows of data representing individuals. The two columns are binned percentages (0-10%, 11-20% etc.), one for two different years. If the responses were numeric, the matched pair analysis would be exactly what I want. But the responses are categorical so that is out.

Any suggestions are greatly appreciated!

Best regards,

Greg

Re: Matched pair analysis with categorical responses

GregMcMahon — Sat, 24 Oct 2020 11:19:58 GMT

Just one other thing to add - it looks like the Agreement statistic may be what I'm after, except I'm more interested in the change and if it is trending up or down.

Re: Matched pair analysis with categorical responses

Mark_Bailey — Sat, 24 Oct 2020 19:16:17 GMT

Agreement between ordinal variables is like correlation between quantitative variables.

The Cochran Armitage test might be helpful to you, too.

Re: Matched pair analysis with categorical responses

GregMcMahon — Sat, 24 Oct 2020 19:44:06 GMT

Thank you Mark!

Re: Matched pair analysis with categorical responses

GregMcMahon — Sat, 24 Oct 2020 19:48:24 GMT

Hi Mark,

For the Cochran-Armitage, are you suggesting I recode the percentage bins to 1,2,3..and make it an ordinal scale?

Thanks again,

Greg

Re: Matched pair analysis with categorical responses

Mark_Bailey — Sat, 24 Oct 2020 19:53:31 GMT

I thought it was already ordinal. The bins as you originally described are find, just change the modeling type to ordinal and make sure that the value orders are correct. If they are not correct, then add the Value Order column property and specify the correct order.