Good evening everyone. If a questionnaire has 4 factors measuring different constructs as for example CAMI questionnaire measuring attitudes toward the mentally ill that has four factors: authoritarism, Benevolence, Social restrictiviness and Community mental health ideology. Is it possible to find good values for cronbach alpha for each factor and a negative value when analysing the whole questionnaire? SPSS says there is a negative covariance. Thanks in advance.
José Luis
Negative Cronbach's Alpha is usually due to reverse-coded items used in the questionnaire during data collection if not coded in opposite directions while measuring alpha reliability. I suggest you to kindly have a look at you dataset.
Hello José Luis Palazón F ,
By theory, as score reliability (in the classical model) is true score variance divided by observed score variance, negative values can't occur. But, since we estimate true score variance as observed (total score) variance minus the sum of the item variances (leaving the covariances, which are unstandardized correlations), if there are sufficient item pairs with negative covariance/correlation signs, then the result can be negative.
The direct implication is that the scores are not very dependable (specifically, that there is not a consistent pattern of performance or response across the set of items), and the measure needs more work. One possible cause of this, as Imran Anwar 's answer notes, is when items are miskeyed or reverse coded, so as to introduce the negative correlations.
Good luck with your work.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS