The first Likert-scale questionnaire we would like to correlate has 28 items and the second one has 10 items. When we used the Pearson-r correlation in the SPSS software, the results would show that it 'cannot be computed because at least one variable is constant'.
CORRELATION calculation required the data to be paired. You have n1 and n2 not of equal size; they cannot be paired. This issue had been addressed by other at RG at the link below.
https://www.researchgate.net/post/How-to-do-correlation-analysis-with-two-variables-in-different-sample-size#:~:text=The%20classic%20correlation%20coefficient%20is,correlation%20coefficient%20of%20both%20variables.
Is there an item for which all participants scored in the same category? That could explain this issue.
I disagree with Paul Louangrath because having different numbers of variables in your scales is not related to having different sample sizes (Ns).
As Christian Geiser suggests, you must have one variable where everyone gave the same response. You can easily check this with descriptive statistics.
Once you sort which variable is a constant, the next thing you need to do is to create scales from your individual LIkert-scored items. Because there have been so many questions on this topic, I have compiled a set of related resources:
https://www.researchgate.net/post/How_can_one_help_people_create_simple_scales_from_Likert-scored_items
Yes, I am in line with the answers of David and Christian G., the concern is Scored-items.
Daryl Jean Astillo ,
David L Morgan made the pertinent observations. You have two variables, the first summarizes 28 items with some form of total and the second summarizes 10.
By the SPSS answer, you are correlating the items and in at least one of them all the answers were the same. These items do not differentiate the respondents at all and do not provide relevant information for the scales. Therefore, they can be removed, and the totals of the scales may be calculated with the items that have different answers. That will solve the problem that SPSS points to.
Another different problem is which type of correlation you should calculate. It is not at all obvious that Pearson's is the right one.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS