I have data set of 470 children (250 male and 220 female) from three different models of school on three different variables. I would like to ask which statistics (Para/Non-Para) are suitable for comparison between the male and female and difference across the three different models and calculate impact of IV on DV.
This depends on how your data are distributed. Non-parametric tests are typically used when your data set is small and/or not normally distributed and therefore doesn't meet the assumptions of a parametric test. If the data meet the assumptions for a parametric test (different tests have different assumptions) then you may have more power with the parametric test.
The kind of statistics to be used depend on the types of your dependent variables and their distributions. What types are your dependent variables?
93
Your data are multilevel data. You have data on 470 children, but these children are clustered within schools. I don't know how many schools you have, but it's bound to be a small number. The mistake is to treat the school variable as if it had been measured 470 times.
Your hypothesis involves an interaction between sex (470 observation) and type of school (a much smaller number of observations).
You need to look at the literature on multilevel modelling. Indeed, many of the original developments in multilevel modelling came from educational assessment, where this kind of question is common. The MLWin website has some very helpful resources. They have a showcase of papers which include many educational research analyses. I'd have a look there for a start.
http://www.bristol.ac.uk/cmm/
- Badges
- Science topic