Little MCAR test for scales or complette survey

Fam te Poel

unread,

Oct 26, 2015, 5:01:53 AM10/26/15

to Missing Data

Hi,

I use paneldata accross four waves involving two constructs. I would like to know whether my data are mcar. When I use this test in spss and include both constructs (2 x 4 scales), the test is significant and data are thus not mcar. However, when I examine missing data patterns per scale, so first the one construct and later the other, bot mcar tests are not significant, indicating data are mcar.

What is the correct procedure in this?

Kind regards,

Fam

Jonathan Bartlett

unread,

Oct 28, 2015, 6:41:28 AM10/28/15

to Missing Data

It could be that missingness in one scale is dependent on the values of the other scale, so that if you test for MCAR just using items from one scale there is no evidence to reject MCAR, but once you also include the items from the other scale, it sees that missingness in items of the first scale is associated with the values of some items in the second scale, and hence it is not MCAR.

Fam te Poel

unread,

Nov 2, 2015, 9:21:31 AM11/2/15

to Missing Data

Dear Jonathan,

thank you for your answer, this makes sense.

Kind regards,
Fam

Op woensdag 28 oktober 2015 11:41:28 UTC+1 schreef Jonathan Bartlett:

Reply all

Reply to author

Forward