1

I'm writing a dynamical script which enables the user to select different scales to analyze their reliability. However, I want adress the overall reliability in some kind of global variable. Since the user might select multiple scales which are not related in any empirical or theoretical way, I'm considering just using the mean of the cronbach's alpha value of the selected scales. Is it legitimate to do so?

Comfort Eagle
  • 243
  • 2
  • 9
  • What have you learned so far -- anything that would make you lean one way or the other? Anything specific that you'd like comments on? – rolando2 May 14 '17 at 21:11

1 Answers1

1

As internal consistense is a correlation the average across the values is not legit.

  1. You should use Fisher's z Transformations on your cronbach's alpha values
  2. Average those z values
  3. Tranform average z back into cronbach's alpha value

For more details please have a look at http://www.tandfonline.com/doi/abs/10.1080/00221309809595548

Gregor Ma
  • 89
  • 4
  • The procedure *per se* [is valid](https://stats.stackexchange.com/questions/221246/such-thing-as-a-weighted-correlation/222107#222107) (btw thanks for the reference, I didn't know this one), but it does not answer the question if averaging Cronbach's alpha's over multiple scales makes any sense (I'd rather argue that it does not). – Tim Jun 19 '17 at 09:22
  • Totally with you, that in most cases it doesn't make sense to average the Cronbach’s alpha across non-related scales. – Gregor Ma Jun 19 '17 at 09:34