Hi everyone,

Is it only me who does not have statistics background to do quantitative research method?

My research will use Structural equation modeling (SEM). I read many material about conducting the SEM but still very confused because they are all different.

- One book says there are 7 steps for the SEM, others say 5 steps. And the names of each step are also a bit different. This is what i am confused.
- I tried to learn SEM by myself, everything from scratch. So i everything that i know is very fragmented. I do not know how to understand it in a systematic way. Could anyone please simplify it?
- In the goodness-of fit indices of SEM. There are Absolute fit, incremental fit and parsimony fit. What if Absolute fit is significant but incremental fit and parsimony fit are not significant? Do we still accept the results? Something like that.
- One step to do the SEM include Model estimation. Why do we have to do it? What is the difference between estimation and goodness of fit?
There are many other questions buzzing around my head now. I am stressful today. When ones get PhD graduation, do they know everything about statistics? My skill in statistics seems zero now and still struggling.

I know my questions are silly but i do not know someone else to ask.

Thank you for reading this

Hi Huubte

Is there a class you can audit, or an online course / set of lectures that is comprehensive and has examples you can follow? Something like that might be helpful. I can understand your frustration though if you are diving into SEM without any stats background. I don't know how to do SEM yet myself, but I have some experience with advanced stats. About the two books differing in their instructions - sometimes stats is more of an art - different authors can have different ways of doing things. I would suggest use the book that you find easiest to understand / follow.

This should be helpful, as it's by the renowned Andy Field who has a knack for making complicated things simple:

If you haven't got it already then I recommend his book "discovering statistics", although I think it is only for if you are using SPSS. You'd need to search and see if he has written anything for the statistical package you are using.

Tudor

There are free resources available through Youtube. I have used SEM for a paper and based on my experiences, you do not have to know everything at all. One of the co-authors on the paper (and supposedly is an expert on SEM in the field of psychology) uses Mackinnon program for SEM modeling. After doing my own readings and interacting with her, what I realized is that even she is not aware about model estimation - other than making sure -- that those are not grossly wrong. Her suggestions are: with more practice and projects, you will get better hang of it. So unless you are a graduate student of Statistics with your thesis fully focusing on SEM, I would not worry about it.at the start. You can try to learn more as you move ahead in the project taking a reverse route from applied to theoretical knowledge. (Excuse for typo's - its my phone not me!!)