Hi, does anyone know how to do a post hoc with nominal variables in SPSS? I have 6 clusters and want to know which ones differ significantly from eachother on variables such as gender. With interval/scale variables this can simply be done by selecting a test such as Tukeys-b under the one-way ANOVA option. Is there no other option for me than to do cross-tabs on cluster 1 with 2,3,4,5 etc (15 seperate crosstab tests) when concerning nominal variables? Any advice would be greatly appreciated!
With nominal variables, if you did "overall" cross-tabs and need to delve further into the root of any differences or significant relationships, then the answer is to do more (smaller) cross-tabs between the areas of interest.
This is addressed in a book entitled "Handbook of Parametric and Nonparametric Statistical Procedures" by Sheskin.
Thanks a lot for your advice! Tomorrow I will go to my uni library and dive into this book. I hope to find out exactly what makes a comparison between to clusters a worthy comparison, and which comparisons can be left behind. In the example of gender, it might thus make sense to only figure out if the clusters with the most opposite percentages differ significantly from each other. But I hope to find this out tomorrow. Thanks loads for your help!
Good luck - you'll find it under the chapter on chi-square and then under post-hocs.
Posting Permissions
|
|
Reply With Quote