I gained a sample of 33 respondents who evaluated 13 dimensions of a topic surveyed.
21-level evaluation scale was used. As I understand, I should use some non-parametric tests like U Mann-Whitney to determine whether these results by each dimension are statistically significant or not. Which test(s) could you recommend to properly assess these results?
The survey issue is a restrictions' severity on human activities within 13 types of zones of protected areas (0 points - mildest restrictions, 20 points - most severe restrictions). Each respondent has been acquainted with these sets established by legislative acts.
I attached xslx file below.
What exactly do you want to test? Whether dimensions are different from each other (average level of restrictions across different dimensions)? Then you should use some kind of a paired-samples test, whether parametric or not. Or whether groups of individuals differ from each other on these dimensions? Then it would be an independent-samples test, parametric or not. There appears to be little variance in your data; coupled with a very small sample size, this is not promising for any statistical analyses, even if you find statistically significant effects by chance...
It should be considered the first case because of too small sample in case of comparing groups of respondents within the sample by age, gender, and education. Unfortunately, there was no opportunity to interview more experts, it is too difficult to find professionals, not just interview the public.
Thank you, Maria K. Pavlova !
OK, then I would personally use a paired-samples t-test, because you have rating scales with many response options (not like usual Likert scales). If you have a reason to expect something like a linear or quadratic trend across these dimensions, you could use repeated measures ANOVA (the dimension should then be numbered in some logical order). This test would have more power than pairwise comparisons of each dimension with each (and the latter would suffer from the problem of multiple significance testing, whereby you get 1 random significant result per 20 tests conducted).
- Badges
- Science topic
- Similar topics
- Agricultural Science
- Soil Science
- Soil Analysis