Timeline for Is ETA a good measure for computing the efect size between an ordinal and a nominal variable?
Current License: CC BY-SA 3.0
3 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Mar 25, 2013 at 11:08 | vote | accept | DL10x | ||
| Mar 21, 2013 at 17:38 | comment | added | ttnphns | +1. To add just few things. 1) In ordinal regression, you can get Nagelkerke's R-square. The sq. root of it is probably the most theoretically consistent measure of association bw a nominal and an ordinal variable, like Eta is for bw a nominal and a scale variable. 2) Another, less puristic approach will be to use still Eta, but after the ordinal variable has been transformed into ranks. We often escape the problem of ordinality simply by ranking (that is, transforming to uniform distribution); this is what we do for example when we compute Spearman correlation. | |
| Mar 21, 2013 at 16:45 | history | answered | Jeremy Miles | CC BY-SA 3.0 |