Suppose you are creating a questionnaire which consists of five separate scales,k which have each been found to demonstrate good psychometric properties. One of these scales is scored on a 9-point Likert scale, two of them on a 7-point Likert scale, one on a 6-point Likert scale, and one on a 5-point Likert scale.
To avoid confusing your respondents and ensuring that they have a relatively consistent response format, is it acceptable to standardise the responses (e.g. transforming the scales so that all of them are scored on 9-point Likert scales)? Or should you leave them in exactly the format they were previously presented in, because this is the format in which they demonstrated good psychometric properties? Is there any consensus in the literature on which strategy is more appropriate?
From a statistical point of view, there is no problem using scales with different numbers of responses categories.
In terms of confusing respondents, I would think in terms of the respondents' familiarity with completing surveys. So, if you are working with a reasonably well educated group of respondents who are likely to have filled out a number of surveys in their lives, then have multiple response formats is unlikely to be a problem.
From the perspective of Rasch models you do not need to assign those "weights" According to model fit each category has specific measures. If you want items to have different probability functions (with or without different formats) you can model it.
Check:
https://www.winsteps.com/winman/models.htm
http://www.winsteps.com/winman/example8.htm
For instructions for Winsteps (excellent low-cost software). Explore winsteps.com and you will find more info.
As Prof Morgan said, considering the respondents' level of understanding is very important. If the target population is use to and have know how of filling the questionnaire, most probably it won't create any problem. However, in case opposite case, I will suggest to use same scale.
Good luck!
Actually, most of the questionnaires include scales that are not equal in its number of steps. This is acceptable in research and statistics because the analyses will be for the sum of items in each scale separately. In addition, as long as the introductory stem for each scale should be written in clear {bold and large font} so the participants can recognize the transition to new scale. You may do piloting to check the smoothness and clarity of your scale. Good luck
I assume when you mentioned "scale", it is referring to a set of questionnaires developed by someone to measure certain variables. You mentioned each has good psychometric properties and testing. Thus, the scales should stay that way. Combining different scales in a survey questionnaires depends on how you design it in the questionnaire. You may separate these scales in different pages, etc., have separate instruction on top, etc, highlight certain differences, etc. I have seen scales originally developed with True/False option expanded to a five-Likert scale to provide respondents with more choice (increase sensitivity). At the end, I guess, the important thing is the psychometric and factor analysis test of each scale or sub-scales (for multi-dimensional scales).
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Applied Statistics
- Survey Design
- Questionnaire