Ok, need some help. I'm making some chapter 4 revisions (I hope to defend in 12 days) after asking for (and receiving) some new analysis from "my stats consultant." At this point, I've paid him $500 flat fee plus an additional $100 for asking for the second set of analysis (which, considering how much face-to-face interaction we've had, is a relative bargain). Here's my dilemma and I'm hoping someone can help: On each of 5 scales, there are items that had to be reverse-coded to run the analysis. Fine. He did that. BUT... now I want to go back and show which ITEMS on the scales had the most agreement. So how can I go backwards? If the scales are 5-Likert (and they are), does subtracting the reverse-coded mean from 5 give me the original mean? Is it really that simple?
Alright. After playing with number by hand, I believe that the answer is that the above solution is WRONG and that if its 5-Likert, I need to subtract the reverse-coded mean from 6!
It seems to me that you want to see the alpha level (if each item is deleted), so that you can determine which subset of items within each scale is most homogeneous (or shares the most variance). Most reliability analysis functions in standard statistical software packages will allow you to check a box for that when running the analysis. It is not as easy as you suggested.
I'm actually wanting to view each item individually now - and part of the reason that I hired a stats consultant is because my version of SPSS has expired.
