Hi.
I have made a 5-point Likert-scale questionnaire and invite 31 participants to complete it. In the questionnaire, I have classified all Likert items into 10 scales (i.e., 8 to be independent variables, and 2 to be dependent variables). Those scales are averaged, so they have numeric values (ranging from 1 to 5).
I would like to explore the scales' correlations and relationships. As for correlations, is Spearman better than Pearson? And as for relationships, I would like to explore whether and how my 8 independent variables predict the 2 dependent variables. Can I use any form of regression analysis?
Dear Kristy.
First, choosing a test require testing distribution of your data. As you know spearman is non parametric and pearson is parametric.
But, your data is categorical, so you should consider it as not normally distributed. Also, in many cases İ saw that categorical data assessed with normality test. İ recommend you to doing a normality test, and if your data is normally distributed, so choose pearson for analyzing relationship.
The Pearson is better than the Spearman.
For regression also your data should be normally distributed. İf it is not, so consider Ln function or log.
Hope this help.
İf you have any question ask without any hesitation.
If you have 2 DV and 8 IV, what kind of regression are you interested in? Would you like to build two separate regression functions or to use a multi-target regression? 31 cases seem to be not sufficient for 8 IVs, so perhaps some of your variables will be excluded from the regression function just because of the lack of the data.
The rule of thumb is number of cases must be 4 times greater than the number of (IV + 1).
I would recommend starting with visual analysis, use scatter plots with the jitter effect to visualize relationships. Then a rank correlation coefficient may be useful.
P.S. As I understand, taking logs won't help here as your variables are not log-normally distributed.
- Badges
- Science topic
- Similar topics
- Clinical Pharmacology
- Dosing