During the process of a questionnaire translation from an original language - after accurate translation - you need to validate it in a target sample in order to confirm similar structure in factorial analysis as well as test Cronbach alphas. In my study I obtain a little bit different structure of Polish questionnaire than the structure of original English one. Is it justified to change the subscale that an item originally belonged to? Is it justified to remove it completely due to redundance (because it decreases Cronbach alpha)?
Thank you.
I would recommend that you maintain all items in the translated questionnaire as was in the original because you want to compare like with like.
Hello Michal,
It's not unusual for different samples (from the same culture and native language) to yield data that don't quite fit exactly to the model that authors of a previous study may have identified. So when a new language and culture is added to the mix, factorial invariance can become even more elusive. So the difference may be sample based, culture base, language based, or some combination.
To answer your question, yes, you can change a subscale. You'd document such changes in your description of the measure, of course. As Albert implies in his answer, such changes do make "apples to apples" comparisons more challenging. If it were just one item from a set of 25, I wouldn't worry too much either way (e.g., retaining vs. removing).
Good luck with your work.
Hi, Michal,
The presumption of scale validity across cultures and language groups is a worthy research question in and of itself, so findings that suggest that a scale's revisions improves its cross-cultural validity can be quite important. My recommendation is always to follow the data. If your data set produces a different factor structure, report that in your findings.
Thank you to all of you for your valuable insight! Could you please provide any examples of questionnaires that was modified in that way to adjust to another language (an item removed or recategorized to another scale)? It would be useful to cite that publication.
Thank you!
I would suggest to check the factor structure of the translated scale through Confirmatory Factor Analysis. If the result is valid, then keep the translated scale without changes.
- Badges
- Science topic
- Similar topics
- Psychology
- Psychometrics