-1

I haven't used R in a few years for surveys but I feel like the code for this should be simple. I have a survey broken down into two portions: Importance (1-7) and Satisfaction (1-7). Each student fills out the survey (about 80 total questions so 160 values per student).

I can easily find the mean and S.D. for each question but I am trying to get the potential p-values, frequency distribution, potential trends, etc for an item analysis report. I have looked for help on google but a lot of the code just doesn't seem to work. Any advice?

Uwe
  • 41,420
  • 11
  • 90
  • 134
Dan Paquette
  • 1
  • 7
  • Are you trying to get the mean and SD of an ordinal variable? – Robin Gertenbach Nov 04 '16 at 15:52
  • I did the basic summary which gave means for all the columns. I attempted svymean and some proportion tables to try to get output but I am just really lost. I used the psych package but again, confused on syntax to use. – Dan Paquette Nov 04 '16 at 15:57
  • Try to give us a [reproducible example](http://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-example) otherwise it's impossible to help. – Robin Gertenbach Nov 04 '16 at 15:58
  • And yes I am trying to get the mean value for each question. For instance its like "Do you like the teacher to student ratio" and each student will say on a scale of 1-7 how important is that for a college and how stratified they are with it at this college. If that makes sense. – Dan Paquette Nov 04 '16 at 15:59
  • @Robin Gertenbach: Yes, getting mean and SD of an ordinal variable is what psychologists do in item analysis in a field known as Classical Test Theory (CTT). There are alternatives (ref. Item Response Theory) but this is what people usually do and consider "proven the test of time". It seems to work for them despite obvious mathematical problems. – Bernhard Nov 04 '16 at 16:01
  • I would show more to the survey/link the entire file but it does have sensitive material in it. I thought there must be some simple way to have it go column by column and find those values and print them out. [link](https://snag.gy/XPDOgW.jpg) – Dan Paquette Nov 04 '16 at 16:04
  • Yes, there are functions like `summary` or `colMeans` or `table` but I still consider it best to use one of the packages I mentioned in my answer. – Bernhard Nov 04 '16 at 16:07
  • Ok I will give that a try and let you know. I am inexperienced in R since it has been awhile since I have really coded in it – Dan Paquette Nov 04 '16 at 16:09

1 Answers1

0

Your question boils down to the search for good code/functions On Classical Test Theory. The best place by far to look for that is in the CRAN task view https://cran.r-project.org/web/views/Psychometrics.html mentioning the packages CTT at https://cran.r-project.org/web/packages/CTT/index.html and psychometric at https://cran.r-project.org/web/packages/psychometric/index.html for item analysis. CRAN Task Views are a great way to find the best packages amongst the huge number of available packages.

Bernhard
  • 4,272
  • 1
  • 13
  • 23