What should I do if I am getting negative results for Cronbach's alpha? Is there anyway to adjust it?
Two main causes:
- Small sample size. Even if the assumptions are met and the reliability is decent, an estimate computed from a particular sample can be negative, just as a sample mean is not equal to the population mean. Somewhat surprisingly, whereas experimental psychologists tend to be obsessed with statistical testing and are conditioned to ask a p-value for everything, psychometrics textbooks generally don't care about tests/confidence intervals for reliability estimates. One reason is that their authors assume that any scale development effort will have at the very least hundreds of observations but there is some sampling variability in reliability estimates anyway.
- Negatively worded items/items with strong negative correlation with the underlying factor.
- Recode negatively worded items
- Get more data (you probably don't have enough)
- Remove some item(s)
- Forget alpha (it has some well-known limitations and chances are that it doesn't tell you what you think it's telling you + what were you planning to do with it anyway?)
See also the references in Reverse scoring when question is stated in a negative fashion