# Methods to change Likert Scale

Suppose I have two variables named importance of the product and satisfaction and I need to do comparison with these two. Variable Importance of product is in Likert scale 3 and Satisfaction is in Likert scale 5. How can I convert the importance of product responses into 5 scale?

Also I would like to have a suggestion for my analysis.. Which analysis method is appropriate for comparing these two variables?

We have the survey responses from the customers about certain features and facilities provided to them. One variable is ‘Satisfaction’ which explains their satisfaction with the features and the other one is ‘Importance’ which explains how important these features are. Responses for the variable ‘satisfaction’ are Strongly Disagree, Disagree, Neutral, Agree and Strongly Agree and Responses for ‘Importance’ are Not at all important, somewhat important and Very important.

You cannot change a 3 point scale to a 5 point scale.

If you want to compare the two, and don't have any idea which is the dependent and independent variable, you could use a chi-square, but that doesn't account for the ordinality. There are tests that do, but they are not as well known. I usually use the Jonckheere-Terpstra test.

• For comparison I am referring the suggested test Jonckheere-Terpstra test. Also I got Gap Analysis as another suggestion. Is it relevant here? – Divya Oct 11 '14 at 19:22

You could map $(1,2,3)$ to $(1,3,5)$. That is easy, as a one-by-one recode or $\text{new} = 2 \times \text{old} - 1$, depending on your software.

I can't see any reason for doing that unless you want to average the 3 point scale and present results comparable to those for a 5-point scale. If you wanted to do that, you could naturally work on the averages rather than the raw data.

It would be a bad idea if the goal was to persuade or fool yourself or anybody else that you really had a 5-point scale, whereas you don't.

@Peter Flom has a very good answer to your more general question.

• (+1) It's funny how Peter's "very good answer" has, until this point, collected no upvotes (but I'm about to add mine). – whuber Oct 10 '14 at 16:51
• I was so focused on fooling with my own answer that I omitted to upvote Peter's until now. – Nick Cox Oct 10 '14 at 16:55
• Yes, I often forget too. :-) Frequently I even forget to upvote the question, even though almost always if I think a question is worth answering it must be worth upvoting. – whuber Oct 10 '14 at 16:58
• @NickCox I have added more details to the question. My ultimate aim is to obtain satisfaction index. I wish to know above mentioned mapping ((1, 2, 3) to 1, 3, 5)) is applicable here? If it’s not applicable, is there any other method to analyse it without converting the Likert scale? – Divya Oct 11 '14 at 19:20