From case-study causal models to testing against a generalised dataset: advice requested to get started
I'm a grad student bridging from the social to a mixed-methods approach. I was hoping that I could get a steer regarding analytical methods as I'm new to both counterfactuals and causal inference and structural equation modeling (SEM) and confirmatory factor analysis. I feel like that I need to apologise to all for my imprecise statistical language. I'm trying....
In one study, I have an analysis of 33 case studies (political science / international relations data) where the author reviewed the historical literature and built a generalised causal model with six independent variables, an intervening variable, and a dependent variable.
I would like to take this generalised model and apply it to a larger, different dataset to test for fit. I am of the opinion that this should be via SEM/Confirmatory Factor Analysis (CFA). I'm not certain, as this is a new technique.
If the identified model seems to work (it may be modified during my own research in the future), I would like to search for this model in the historical dataset and make projections concerning the probability of future iterations of the dependent variable.
It is my basic understanding that SEM/CFA is not a tool to make arguments to support causation, but perhaps counterfactuals and causal graphs may be more of an appropriate tool(s).
I'm looking for a basic steer here to get me going down a particular path. I have one graduate level stats course under my belt and am a Stata 15 user.
Respectfully,
John
In one study, I have an analysis of 33 case studies (political science / international relations data) where the author reviewed the historical literature and built a generalised causal model with six independent variables, an intervening variable, and a dependent variable.
I would like to take this generalised model and apply it to a larger, different dataset to test for fit. I am of the opinion that this should be via SEM/Confirmatory Factor Analysis (CFA). I'm not certain, as this is a new technique.
If the identified model seems to work (it may be modified during my own research in the future), I would like to search for this model in the historical dataset and make projections concerning the probability of future iterations of the dependent variable.
It is my basic understanding that SEM/CFA is not a tool to make arguments to support causation, but perhaps counterfactuals and causal graphs may be more of an appropriate tool(s).
I'm looking for a basic steer here to get me going down a particular path. I have one graduate level stats course under my belt and am a Stata 15 user.
Respectfully,
John
