Mine is multiple choice science questionnaire. The correct answer is marked '1' and wrong '0'. The aim of the test was to find out what they know and what they don't know (pre test). The alpha is negative. How to interpret? Should I give easy questions that 60% of the students can answer? In that case how to identify the weak spots where intervention will be focussed?
2
- $\begingroup$ First thing is to look at the inter-item correlations. If many are negative you know where to look next. $\endgroup$mdewey– mdewey2017-09-09 10:49:47 +00:00Commented Sep 9, 2017 at 10:49
- $\begingroup$ I'd also check the between-student variance. If this is too low, then high correlations will not be possible. By the way, Cronbach's alpha for dichotomous data is the same as Kuder-Richardson Formula 20 (KR-20). $\endgroup$Jeffrey Girard– Jeffrey Girard2017-10-17 23:04:17 +00:00Commented Oct 17, 2017 at 23:04
Add a comment |
1 Answer
$\begingroup$ $\endgroup$
Cronbach's alpha isn't really about the difficulty of the items, although it will be low if all the questions are so hard that the answers are random. So, if you asked questions about calculus to 1st grade students, you'd get a very low alpha - maybe even negative by chance. The same happens if the questions are all ridiculously hard.
A very low alpha is saying that the interitem correlations are low. There are a couple reasons this could happen. If you are asking about different areas of science, then you would expect that. If some of the answers were coded incorrectly, you can fix that.
So, look at the inter-item correlations and, especially, look at correlations that are negative and think about whether that could be expected.
