Friday, February 24, 2006
st: thoughts on an analysis plan?
We have ratings of skill demands on many dimensions (let's just say it's a lot) for approximately 50 jobs. We have three hypotheses we'd like to test against each other. First, that all the jobs have the same skill demands (i.e., all the jobs are the same). Second, that there are really three groups of jobs, where the jobs in each group are more similar than different on these skill demands. These three groups are identified a priori. Third, that it's better to identify job groupings empirically than a priori.
To begin, we will probably reduce the y vector by using something like principal components or factor analysis. This we are quite clear about.
Next, we're thinking that we can basically get R-squares for each hypothesis that we can compare.
1) For the R-square that all jobs are the same, we will use coefficient alpha. 2) For the three-group models, we were hoping to use an R-square (or pseudo R-square) from MANOVA.
However, Stata's MANOVA doesn't report R-square, but does give Hotelling's T-square... This has led me to consider whether we should try to compare F-statistics across these three models...
* * For searches and help try: * http://www.stata.com/support/faqs/res/findit.html * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/
Links to this post: