Dear all,
In a risk perception survey, I would like to calculate the Cronbach's alpha for this type of question: do you think this environmental factor is at risk?
- Strongly disagree (value=1)
- Disagree (value=2)
- Neither agree nor disagree (value=3=
- Agree (value=4)
- Strongly agree (value=5)
- I don't know ----> I don't know how to code this answer. I think it may be coded -10 or +10 or +100 as value because it is out of the scale from 1 to 5. However I don't know if it right. 1) How would you code this value? Or should I delete this observation in this dataset?
2) Do you think I should gather the values "Strongly diagree/disagree" and "Agree/strongly agree" together, if my number of answer is low?
Thanks in advance,
Kind regards,
Thibault
I would only conduct a reliability analysis on the items on the scale 1-5. You can state descriptively the % of people who endorsed "I don't Know." Also, no, I would not combine items before conducting the analysis. If your Cronbach's alpha is low, check if there are any individual items contributing to that. It could also be a dimensionality problem (run a quick EFA or PCA to see how many factors are indicated). Also, check to ensure that there are no items that need to be reverse coded because if so, and you did not do that (or did it incorrectly), it'll be reflected in the reliability coefficient. These are places I'd start.
Cory
- Badges
- Science topic