Hi guys,
I need help... My PhD thesis is about consumer behaviour analysis. In detail I look at what actions by companies trigger what customer reactions. In my quantitative study, I would like to look at the following dependent variables:
Behavior (with respective items)
Attitude (...)
Emotion (...)
And the following independent variables:
Opportunity for explanation (given/not given)
Alternative presented (yes/no)
My question is... Do I really need SEM? My professor wants me to do so, but I would rather like to use an experimental MANOVA design.
I would appreciate any help!
Printer-friendly version
MANOVA, regression, and SEM are all special cases of the general linear model. ANOVA and regression give the same answers when provided the same data, as to MANCOVA and identically specified multi-level models/heirarchical linear models/SEMs.
The benefit to using SEM like methods is how general and flexible they are. MANOVA/MANCOVA methods will have restrictions on how the set of dependent variables are related/treated. A basic model where Opportunity and Alt. Present predict Behavior, Attitude, and Emotion will fit equally well in either framework. If you want to test constraints on any parameters, include individuals who have missing data under full information methods, or include item level behavior variables in a factor model, SEM allows you a lot more flexibility.
It's your thesis. You've got to use something you're comfortable with and can defend. If you can answer your question using either package, it's your choice. However, its possible that your professor is suggesting SEM because s/he has questions that can only be answered in SEM. A dissertation is a good time to learn a new method, as it can allow you to answer a big question. Good luck!
Ryne, thank you so much for your detailed answer..! I really appreciate it.
I hope it is okay, that I have one additional question...
For an academic, how difficult is it to learn SEM? I have read some papers on this method and googled all kinds of information - many sources say it is highly complex, I even read "nobody reallly understands SEM". That is why I am a bit intimidated. It is possible to learn it from books or would you say this method definetly requires anyone to visit a seminar? My PhD is a distance learning program, so unfortunately I do not have statistic lecturers at hand.
Thank you so much!
Christoph
We at OpenMx think SEM is pretty accessible, but it will depend on your interest, existing knowledge, and how adept and willing you are utilize whatever instructional resources at your disposal. Steve Boker's 2010 SEM class is on our forums (http://openmx.psyc.virginia.edu/forums/openmx-help/teaching-sem-using-openmx/uva-introduction-sem-spring-2010), as are resources from some of our previous workshops. Other packages have their own online resources, including lavaan (http://lavaan.ugent.be/tutorial/) another free R SEM program we "compete" with. In person workshops commonly occur at academic conferences and as stand alone programs lasting anywhere from half a day to a week.
There are plenty of good books and online resources on our "Resources" page (http://openmx.psyc.virginia.edu/sem-resources). If your professor is recommending this, see if s/he can help you find support for this!