What is the tolerance for missing values in the construction of SEM

[Elias Carvalho]

I would like to know your opinion about how much missing data I can consider before apply a multiple imputation when building models with lavaan.

For example, what is the acceptable amount of missing data? 10%, 20%, 50%

Send me literature if possible please.

[Jeremy Miles]

It's not missing data that you care about, it's missing information.

If you're missing 90% of cases on a particular variable, but you're able to explain 99% of the variance in that variable using other variables, then you're not missing very much information. If you're missing the group to which people were randomized, there's probably no way to recover that, so you're missing a lot of information

My favorite paper on this is by Savalei and Rhemtulla: https://www.tandfonline.com/doi/abs/10.1080/10705511.2012.687669 

--
You received this message because you are subscribed to the Google Groups "lavaan" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lavaan+un...@googlegroups.com.
To post to this group, send email to lav...@googlegroups.com.
Visit this group at https://groups.google.com/group/lavaan.
For more options, visit https://groups.google.com/d/optout.

[Terrence Jorgensen]

You can ask a general question like this on SEMNET

http://www2.gsu.edu/~mkteer/semnet.html

In my experience, the amount of missing data is not as relevant as how much information about the parameters is missing (the "fraction missing information").  Enders (2010) book "Applied missing data analysis" is a great text to read.

Terrence D. Jorgensen

Postdoctoral Researcher, Methods and Statistics

Research Institute for Child Development and Education, the University of Amsterdam

UvA web page: http://www.uva.nl/profile/t.d.jorgensen