Kappa w/ unequal score ranges in SPSS 19

[jnk7711]

Hi all -

I just upgraded to SPSS 19 and I went to run some syntax I wrote in SPSS 18. In SPSS 18, I was unable to compute kappa when my raters had unequal score ranges - that is, when the matrix for two raters was not symmetrical. For example, if rater 1 used scores of 1, 2, 3, 4, and 5 but rater 2 only used scores of 2, 3, 4, and 5 then SPSS 18 would not compute kappa. To compute kappa, I added in dummy variables to create a symmetrical matrix.

It appears that in SPSS 19 this is not needed. That is, it appears that in version 19 unequal score ranges are no longer an issue, and SPSS will automatically correct for a non-symmetrical matrix.

I was hoping that someone else can confirm this for me - I have been having trouble finding out whether or not this is the case.

[Bruce Weaver]

The way to fix the non-square table problem is (or at least was) to use the integer mode for CROSSTABS.  E.g., if there are 5 possible ratings (1-5), and r1 and r2 are the two raters:

CROSSTABS  VARIABLES=r1 (1,5) r2 (1,5)
  /TABLES=r1 by r2 / stat = kappa.

I don't have v19 yet, so cannot comment on what it is doing.  What does your CROSSTABS syntax look like?

HTH.

jnk7711 wrote

Hi all -

I just upgraded to SPSS 19 and I went to run some syntax I wrote in SPSS 18. In SPSS 18, I was unable to compute kappa when my raters had unequal score ranges - that is, when the matrix for two raters was not symmetrical. For example, if rater 1 used scores of 1, 2, 3, 4, and 5 but rater 2 only used scores of 2, 3, 4, and 5 then SPSS 18 would not compute kappa. To compute kappa, I added in dummy variables to create a symmetrical matrix.

It appears that in SPSS 19 this is not needed. That is, it appears that in version 19 unequal score ranges are no longer an issue, and SPSS will automatically correct for a non-symmetrical matrix.

I was hoping that someone else can confirm this for me - I have been having trouble finding out whether or not this is the case.