Currently i am doing an accounting research in a hospital entitled "The impact of authority structure on budgetary control" in one public hospital. What i want to know here is that, i am planning to use purposive sampling. That means, i will purposly select clinical managers from the hospital so as to see the impact of the formal and informal authority that they have on controlling their behavior using budget. I measure their formal and informal authority , as well as my dependent variable i.e the budgetary control using five point likert scale. What i want to know here is that is it possible to use multiple regression analysis to study the impact of authority structure on budgetary control using PURPOSIVE SAMPLING in just a single organization? Thank you .
Are you picking out specific clinical managers based on your criteria. If you're using all managers, or a subset selected randomly, then it shouldn't be a problem.
What sort of regression are you using? For a Likert scale, I'd consider either ordinal regression or logistic regression using a threshold. Linear regression is going to assume that each step of the 5 point scale is equal, and that the outcome isn't bounded.
i didn't get both of you guys, please make it clear. I am gonna use standard multiple regression.
Mitchell point is about the data. A purposive sample would mean you deliberately selected a subset of managers from the list of all such managers. This would be a non-probability sample for which you cannot run any regression analysis. All managers = the population or a random sample from the population list would be a probability sample thus ok for a regression analysis.
I also agree with Mitchell on the type of regression. Likert statements are ordinal level since they are based on word choices. The"numbers" are just codes to simplify analysis. They are not exact values. In addition, "standard" multiple regression implies linear regression which requires the dependent variable to be interval or ratio level. A Likert index is not interval or ratio since it is based on ordinal level data.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Applied Statistics
- Survey Design
- Questionnaire