Fully conditional specification (MCMC)

[rastiola]

I have a small problem with "impute missing values":

I found it hard to find the full algorithm of spss to this method. If the numeral form of my data is: age, education, own apartment/not, salary etc. There are 50000 rows of information and about 5% of them are missing arbitrary.

Fully conditional specification-method seems to work pretty well here. But: For example, if there are missing value in education. I wonder if the method ignore that if person have good salary, he/she probably must have own apartment and good education. So is it safety to use this method or shoud i rather use series mean for example?

And sorry for my bad english.

[rastiola]

.........Anyone?