How I would calculate the correlation between 9 categorical binary variablçes (0,1) and 12 categorical ordered variables (1-3, 1-5) using R?
Asked
Active
Viewed 199 times
0
-
3Out of concern that this could be arbitrary and meaningless, I wouldn't do it. I would stop and ask myself why I am even proposing such a procedure; I would consider what my actual underlying objective is; and I would inquire whether there might be a different procedure that would achieve my aims. – whuber Nov 10 '17 at 17:17
-
1I'd suggest using some kind of graphical tool, like a heatmap to understand the relation between categorical variables (a simple table could be enough with a small number of categories). – Freguglia Nov 10 '17 at 17:25
-
For each pair of binary and ordered variables, you could use Cochran-Armitage test for the test of association and either Freeman's theta or epsilon squared for the degree of association. – Sal Mangiafico Nov 10 '17 at 19:11
-
Also, spineplot in the coin package is useful for visualization. – Sal Mangiafico Nov 10 '17 at 19:12