The survey consists of 4 sections.
Demographics
A frequency scale (17 items)
A likert scale (15 items)
6 subscales of 10 items each.
I administered the survey to two classes . Resulting in four types of respondent
All respondents,
Group A
Group B
Those who responded “Other”
The first problem: the one group is way, way too small.
Population A: 589
Population B: 388
The second problem is that several participants dropped-out after each section of the survey.
I think the best way to address the issues are to run the same statistics on all the types of respondent, and ignore the small sample size of the one group.
I’m open to alternative suggestions.
The first thing I need is item-total correlations for each item.
The biggest problem I have is in relation to my reliabilities.
I wrote the 60 items, 10 items per construct for 6 constructs
One of the papers I used did something similar, they used too many items, then removed some after ata collection to improve reliabilities, so I would like the same thing done to my subscales.
Once the scales are reliable, I need a CFA.
Then correlations and regressions on the reliable versions of the scales.