I'm doing some data processing, and I need to check the origin of my missing data before I can start multiple imputation.

Unfortunately the BaylorEdpsych and the MissMech package are not available on cran anymore and the naniar package does not do the mcar_test.

Has anyone got an alternative package to do Little's mcar_test? Maybe even a better solution how to check for that?

I know that testing if data is MNAR is almost not possible. This is the first time I am doing this and in theory I would argue that the data missing at random. But as there are a few variables with almost 50% missing values, I guess I need statistical proof to determine that they are unlikely to be "MNAR" to justify the multiple imputation procedure.

1

There are 1 answers

0
Nick Tierney On

naniar now has mcar_test thanks to a pull request from Andrew Heiss!

http://naniar.njtierney.com/reference/mcar_test.html?q=mcar