For testing reliability/statistic work on data, all the questions for every construct (for example attitude) should be in the same direction in terms of being agree or disagree? If the questions are not in the same direction, I should transfer value before SPSS testing?
You must do transformation in order to ensure the same direction. I have found that people will argue items should all be positively associated with each other.
You should consider removing it too. If the coefficient is small, it can be explained as a misinterpretation.
Dear Marjan Jafarpour, There are scope of modifying your input data according to your need. But a small MACROS can be designed.
The most valid approach is to keep the scoring of items in the same direction (i.e., lowest to highest, 1 to 5, etc.) That lowers the negative instrumentation effect threat to internal validity.
Yes, you need to transform your raw data in SPSS. You can do this by subtracting the original score from the highest value + 1. So, if you are working with a five-point scale, subtract from six (6-5 = 1; 6-4 = 2, etc.).
If your items are working as expected, you should have nothing but positive correlations. Also, when you run the Reliabilities command, you should include the option to get values for "Alpha if item deleted." That will show you if there are any cases where the reliability would go up if you dropped an item from the scale (which would definitely be the case if an item were negatively correlated with the other components in the scale).
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software