I am using a self-designed questionnaire to assess environmental affect in young learners. The items in my questionnaire refer to different sub-constructs within the environmental affect domain, such as attitude, emotion, concern, and locus of control. However, the questionnaire was not specifically designed to function as a scale, and I do not intend to analyze the sub-constructs as separate scales. I have gathered pre-test and post-test data for each item in my questionnaire. I plan to use paired sample t-tests to compare the means of individual items (pre-and post-test scores) to assess changes over time. Given that my focus is on comparing individual items rather than analyzing the overall sub-constructs, I am wondering if it is necessary or appropriate to report Cronbach's alpha for this type of item-level analysis. Are there other reliability measures I should consider, or can I proceed with the paired sample t-tests without needing to establish internal consistency for the questionnaire as a whole?
The answer to your question is in your statement: The questionnaire was not intended to function as a scale. If it's not a scale, then there is no reason to assess the items at the scale level--that is, don't use Cronbach's alpha. Here, however, is a caveat: just because you didn't intend for the questionnaire to function as a scale doesn't mean that your respondents did not reply as though it was a scale. At the very least, you should look at inter-item correlations. If two or more items are highly correlated, then the t tests you plan to calculate will not be independent. When analyzing your data, you are trying to find out how the respondents behaved, which should not be based only on your intentions.
- Badges
- Science topic
- Similar topics
- Biological Science
- Bioecology
- Systems Ecology
- Landscape Ecology
- Connectivity