Missing data mechanisms

[@DontKillAnts]

Great explanation. I am so surprised at poor pedagogy of the published article I'm reading talking about the differences between these mechanisms. You provided a clear explanation with a good (albeit borrowed) example. Good job.

[@mronkko]

Thanks. My example comes from Enders' book, which in my opinion is pretty much the only thing that you should read about missing data.

[@johnspivack]

Very good, very helpful. Thank you.

[@mronkko]

You are welcome!

[@DOLCEKAYEXOTICAL]

Thank you for such a straightforward and simple explanation

[@mronkko]

You're very welcome!

[@mersaultjude]

Thank you!

[@mronkko]

You're welcome!

[@haraldurkarlsson1147]

Is the MNAR case with job performance a case of survivor bias?

[@mronkko]

Depends on the context. If you have missing data in job performance because poor performers are fired permanently but you still somehow have data on at least some of the variables for these individuals, then yes.

[@haraldurkarlsson1147]

@@mronkko As a physical scientist I find the subject fasinating but complex.

[@PavelBalov]

Hi! Could you please advice, how to use Little's MCAR test in python? It is easily implemented in R, but there is no popular package with MCAR test in python. Maybe Little's test can be manually coded as a sequence of other statistical tests?

[@mronkko]

I do not use Python, so I cannot help you with that specifically. However, implementing the test is generally not difficult. For example Enders (missing data book, 2010) gives the formula on page 20. You just implement that and compare that against the appropriate chi2 distribution to get a p value.