I want to analyze the relationship between a student's experience of bullying and other variables using regression, the Process macro, or SEM.
1. Should I recruit both students who have experienced bullying and those who have not, or should I only recruit students who have experienced bullying?
2. If I need to recruit students who have not experienced bullying, should I control for the proportion of students who have and have not experienced bullying? I know that in regression analysis, the assumption is normal distribution for variable, but if I have a higher number of students who have not experienced bullying(Which will be coded as 0), won't the distribution be skewed instead of normal?
1) This depends on how you want to frame your research question, if you want to specifically only check students' experiences of bullying and certain variables, you only need to recruit students who experienced bullying. Another take on this could be to create two models, one for students who didn't experience bullying and one model for those who did, and ask "are these models any different?" using Chi square or whatnot (or create a nested model with "was bullied"/"not bullied" as a categorical factor/level)
2) There's a small but important error here, "I know that in regression analysis, the assumption is normal distribution for variable", this is not the case at all, the variables doesn't have to follow a normal distribution, it's the residuals (i.e., errors - the mismatch between your observed and predicted values) that have to approximately follow a normal distribution.
- Badges
- Science topic