# Thread: Separating 5 item construct with scale 1-7 in three groups

1. ## Separating 5 item construct with scale 1-7 in three groups

Hi there!
I would like to separate my 5 item questionnaire construct (with a scale from 1 to 7) in three groups (1 (low) - 2 (middle) - 3 (high)) to investigate group differences.
Can anybody help me and tell me how I could do that?
Does somebody remember a paper where somebody has done this before?

2. ## Re: Separating 5 item construct with scale 1-7 in three groups

I would like to separate my 5 item questionnaire construct (with a scale from 1 to 7) in three groups (1 (low) - 2 (middle) - 3 (high)) to investigate group differences.
Why? Why not just use the sum score? Categorization of continuous data is usually a very bad practice. It destroys statistical information, and it can produce misleading results. Maybe you could provide a bit more context (what is your study about, what is your research question, how was the study designed, where does your questionnaire come from etc.)?

With kind regards

Karabiner

3. ## Re: Separating 5 item construct with scale 1-7 in three groups

Originally Posted by Karabiner
Why? Why not just use the sum score? Categorization of continuous data is usually a very bad practice. It destroys statistical information, and it can produce misleading results. Maybe you could provide a bit more context (what is your study about, what is your research question, how was the study designed, where does your questionnaire come from etc.)?

With kind regards

Karabiner
Thanks for asking! My study is about differences in behavioral norms of customers. The customers were asked in a questionnaire how they regularly behave when cooperating with suppliers. The data shows that the are two modes (one at 1= totally disagree and the other at 4 = neither agree nor disagree). Therefore the distribution is non-normal. I would assume there are 2 (or maybe 3) different groups who behave in totally different manner.

 Tweet