Question on Whether should i use SEM on Sociological Data Analysis?

[rowena chen]

Hi dears,


I am a new member on this forum and i'm very happy to find a good place asking questions.

I am a phd candidate in Sociology and my topic is to explore the relationship between two big variables (i'm going to hide some details), simply speaking, the relationship between someone's individual competency (16 items) and his/her working effectiveness (10 items).


I use survey to get the data that i want (closed-ended questions + a few open-ended questions)


- closed-ended questions consists of a couple of 5-likert questions (measured through two mature scales)

Because the open-ended question part do not limit the minimum word count, so the word count that participants returned to me is between 50 - 150, so i don't think this part can be qualitatively analysed, therefore i regard my study as a pure quantitative study, not a mixed method study. (any other ideas please feel free to shout out)



My questions is, because my research includes latent variables (competency, effectiveness), therefore i guess SEM could be a good tool to analyse data (the reason why i say'guess' is because i do qualitative study more in the past, i want to know more about quantitative one before i meet with my supervisors).


what do you think? many thanks.



Best wishes,
Rowena

[Richard Williams]

I don't quite understand the question. I'm a sociologist and I'm not sure what "Sociological Data Analysis" is. Sociologists use many methods, SEM being among them.

For your open ended Qs,you might code them so they can be used in a content analysis. But a lot of times, all those quotes are used to "humanize" the numbers from the rest of the analysis,

You might use SEM. You might use factor analysis. I can't really tell you what to do but I suspect you should read up on the methods first. The Stata sem model is very good and has many good examples, Acock's book on SEM is also very good:

https://www.stata.com/bookstore/disc...g-using-stata/

[Mike Lacy]

To the good advice offered by my fellow sociologist Richard Williams I'd add the thought that your problem is as much substantive/theoretical as it is statistical. I'm sympathetic to the idea that sometimes statistical method choices can give us different ways of thinking about the substance of a problem, but I'm more inclined to the view that, if you have detailed substantive ideas about the kinds of questions you'd like to ask of your data, the choice of statistical method will be clear(er). Of course, there also can be an iterative process here, where the method informs the substance, which in turn affects the method, which in turn ... . I'd also say that looking at methodological approaches in previous sociological (or other) research on the topic would be a good way to decide about whether SEM or some other statistical method would be useful.

[Richard Williams]

When I was in graduate school, one of my fellow students asked what method they should use. I told them Structural Equation Modeling seemed good, and they responded "I'm writing a substantive dissertation, not a quantitative one." Which struck me as odd, because I had never thought there was a conflict between the two: You use the method that is appropriate for addressing your topic.

Conversely, another student said they wanted to use SEM, and I said it didn't seem necessary given what they were doing. They responded that they "wanted employers to be impressed by their quantitative skills." I thought well, ok, but I suspect employers might be even more impressed if they used SEM for a topic that was best analyzed with SEM, instead of for a topic where SEM was unnecessarily complicated.

Anyway, as Mike says, I don't think it is best to decide on the method before you know more about the substance of your problem. If you don't have much quantitative background, jumping straight to a more advanced method like SEM probably isn't the best idea. Learn more about SEM and other methods first and think more about the questions you want to address before deciding what methods you are going to use.