Imputing Data for a Case w/ No Values for a Given Measure

#1
Hi, all, I have a measure for which I have two cases missing values for all items. The missingness is MCAR (due to experimenter error); is it possible (and more specifically, to any degree methodologically sound) to impute values for these cases' items using observations from other available cases? :confused:

Thank you very much,
 

trinker

ggplot2orBust
#2
Yes you can do this. I've seen a method that basically is like linear regression that uses what information you do have and the information of similar rows (observations) to predict the missing values. However in a case with MCAR and just 2 I'd say do a list-wise deletion and explain this in the write-up unless sample size is a problem.

My thoughts but others may weigh in here.
 

Lazar

Phineas Packard
#3
If you have no data on these cases whatsoever then no you cannot impute as you have no inputs into your missing data model. However, if you have some data (age of cases, demographics, etc.) then yes you can impute.

In any case what TR says is correct if the data because your data are MCAR listwise deletion is unbiased.