# Creating a summative scale for items with different number of response categories?

#### menteith

##### New Member
Dear TalkStats users,

Let me first give you my main goal: I would like to create a summative scale that I could compare with other scales (each of these two is Likert's scale). However, I see a large obstacle to creating a summative scale: the items have a different number of response categories. From what I have learnt so fair it seems an insurmountable obstacle.

However, let's go into some more details about a summative scale I'd like to create. I use a survey I was given to. People from four different groups were surveyed. From the survey I have chosen five questions (items), which I think (based on the literature and previous research) measure democracy in small groups. For the sake of clarity, let me first describe one item for the rest of them give only possible responses.

Q1. How do you estimate the changes of the propositions that are put forward by ordinary group members to be taken into account?

1. they are very probable
2. they are average
3. they are very slim
4. I don't know

Q2.

1. all
2. majority
3. large group
4. elite
5. one person
6. I don't know

Q3.

1. they have strong influence
2. they have limited influence
3. they have no influence
4. I don't know

Q4.

1. they have strong control
2. they have limited control
3. they have no control
4. I don't know

Q5.

1. strongly disagree
2. disagree
3. agree
4. I don't know

First problem. As you can see the possible answers to the items varies: it is 4 in 4 out of 5 cases; in Q2 it is 6 (this question is quite important). Perhaps responses to the Q5 are the most similar (compared to other responses) to that of Likert's scale.

Second problem.
I find it hart to choose the right response for each question (item). At the outset I assumed that there would be a great deal of variation between "extreme" responses (that is: first and last). However, this is not the case. Please see the responses to Q4:

Group A:
3,8%

Group B:
4,2%

Group C:
15,5%

Group D:
3,75%

With the exception of Group C, the percentages of responses within each group are quite similar. This comes as no surprise: participants avoided extreme response categories.

When it comes the response category to the question Q4 they are as follows:

Group A: 25,0%
Group B: 23,6%
Group C: 32,7%
Group D: 44,1%

I think this response category number 2 tells more that "extreme" responses.

There is another difficulty thought. Please the following example (fictional but in my database there are such responses):

Group A:
Q1: