I estimated a CFA in R using missing = "ML" in model syntax. Does this "ML" do something about the no-response bias? Or no-response bias is a totally different thing than dealing with missing observation in a survey study?
Thank you
If there are unanswered questions you can delete those evaluations so you can perform your analysis.
Hello Ayesha,
You didn't specify which R library you're using to execute the CFA (confirmatory factor analysis). But, for any R library/package, you can always learn about the features and functions by typing "help(package name)"
For example, if you were using lavaan, the command would be help(lavaan)
Good luck with your work.
Ayesha,
most (if not all) CFA packages should be able to handle non-response via maximum likelihood estimation or multiple imputation. This does depend on the missingness pattern to some extent (see the paper cited below).
I would not recommend deleting cases, as that reduces power, wastes information, and might introduce bias.
See: Article Missing Data
- Badges
- Science topic
- Similar topics
- Statistics
- Survey