1

I have three questions in a survey that are supposed to measure different constructs. However, I have my doubts if the respondents will not confuse them. So, I know that Cronbach's $\alpha$ is used to check if different questions pertain to the same construct, but I wonder if it is reasonable to use it in the opposite way. Thus, if Cronbach's $\alpha$ returns something below 0.6 (poor) I could consider that those questions do not pertain to the same construct.

The questions are ordinal Likert scale as follows:

How do you think it was to find the information in the briefing?
(5) Very easy
(4) Easy
(3) Neutral
(2) Difficult
(1) Very difficult

Is the briefing interface clear and understandable?
(5) Strongly agree
(4) Agree
(3)Neutral
(2) Disagree
(1) Strongly disagree

Does the briefing look reliable?
(5) Strongly agree
(4) Agree
(3)Neutral
(2) Disagree
(1) Strongly disagree
Stephan Kolassa
  • 95,027
  • 13
  • 197
  • 357
bfsc
  • 11
  • 1
  • I'm pretty sure Cronbach's alpha is *not* "used to check if different questions pertain to the same construct" (or at least shouldn't be). You use factor analysis for that. – gung - Reinstate Monica Mar 04 '16 at 00:44
  • Strange @gung if you look in wikipedia (http://bit.ly/1QWArNI) they say that it is. I know wikipedia is not very reliable, but I read some serious research papers that use it like this. – bfsc Mar 04 '16 at 00:55
  • I don't doubt (unfortunately) that Cronbach's alpha is used to assess unidimensionality. Nevertheless, that is incorrect (cf, [Assessing reliability of a questionnaire: dimensionality, problematic items, and whether to use alpha, lambda6 or some other index?](http://stats.stackexchange.com/a/11846/7290)). – gung - Reinstate Monica Mar 04 '16 at 01:26

0 Answers0