Dear All,
I need your help :)
For my thesis, I've done an experiment and I need to analyze it on SPSS.
Specifically, I have 112 participants who had to read 7 different scenarios, and after reading the scenario, they had to indicate their degree of agreement on an ascetic scale with five different statements that were intended to measure my dependent variables. The aim is to analyze how the different scenarios impact the dependent variables.
independent variables: scenario 1, scenario 2, scenario 3, scenario 4, scenario 5, scenario 6, and scenario 7
dependent variables: motivation, satisfaction, collaboration and help.
control varibale : gender( male, female ), age and status ( employed, unemployed, student , self-employed, retired, other )
My supervisor advised me to do a multiple linear regression that took into account the clustered standard error ( in order to control for individual heterogeneity). He explained that this would take into account my design within subject.
Unfortunately, I'm a beginner in statistics and this is the first time I've used Spss, so I can't figure out how I could permofer a test that would take clustered standard error into account.
Could you please help me and explain how to do it on spss?
You mention a supervisor. Questions like this ought to be asked with supervisors. Otherwise you may get into a situation where you've used a method, perhaps devoting a lot of time to it, and then your supervisor says it is wrong and you've wasted your time. If your supervisor doesn't know how to do something you've been asked to do, ask other professors in your department. If you find no expertise in this way, ask your supervisor who to contact, or what resources might be helpful. If you are in a decent university, there will be a library. It would not be a bad idea to go there and spend a few days looking at all the statistics books.
I agree with Dan E Kelley that it is best to discuss these issues with your advisor. Nonetheless, here are some thoughts (again, discuss them with your advisor before you decide what to do!):
You could use multilevel (hierarchical linear) modeling, but with only 7 clusters (scenarios, level-2 units), this may not be necessary or fruitful. Instead, you could use the (simpler) "fixed effects approach to clustering" in which you dummy-code the scenarios (using 6 dummy code variables such that one scenario serves as reference/comparison standard) and use ordinary least squares (OLS) regression (i.e., standard single-level multilevel linear regression analysis). Both approaches are described, for example, in
Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2013). Applied multiple regression/correlation analysis for the behavioral sciences. Routledge. (Chapter 14)
- Badges
- Science topic
- Similar topics
- Psychology