Choose Language Hide Translation Bar
Highlighted
jvh
jvh
Level I

How do I perform a nonparametric test for Likert data?

Dear experts, 

 

Pardon my novice question. I am new to JMP Pro (14.2) and to statistics. I have three survey questions, each on a 5 point Likert scale, which were answered by two groups, group A and group B. There is a different N between the groups. I would like to assess for any difference, on each question between the two groups (e.g. question 1 responses from A compared to B). 

I had planned on a Mann-Whitney U test. I put in Likert responses in as ordinal (see attached image). When I Analyze Y by X, I do not see the nonparametric option. I only see nonparametric if I change the Data Type to continuous.  

I am able to complete a Mann-Whitney U in JMP Pro? Do have to use the Wilcoxon test? If so, is it ok to consider the Likert data interval, not ordinal, in this case? 

 

Cordially, 

Jared 

 

 

1 ACCEPTED SOLUTION

Accepted Solutions
Highlighted

Re: How do I perform a nonparametric test for Likert data?

Nonparametric techniques work on ranked data. The measurements are usually continuous, then ranking is performed and the procedure would be performed on the rankings. That is why JMP is expecting a continuous response, not ordinal. Even if the data were already ranked, JMP would "re-rank" resulting in the same data. 

 

If you truly want to use a nonparametric technique, make the column continuous. Then you can perform the technique. Keep in mind that Likert data does not necessarily require a nonparametric technique, but this is a point of discussion among statisticians.

Dan Obermiller

View solution in original post

1 REPLY 1
Highlighted

Re: How do I perform a nonparametric test for Likert data?

Nonparametric techniques work on ranked data. The measurements are usually continuous, then ranking is performed and the procedure would be performed on the rankings. That is why JMP is expecting a continuous response, not ordinal. Even if the data were already ranked, JMP would "re-rank" resulting in the same data. 

 

If you truly want to use a nonparametric technique, make the column continuous. Then you can perform the technique. Keep in mind that Likert data does not necessarily require a nonparametric technique, but this is a point of discussion among statisticians.

Dan Obermiller

View solution in original post

Article Labels