One possibility is transforming the values of your categories into
the midpoint value of each income bracket, with a view to assigning the
midpoint to missing cases.
If your predictors are of an objective nature (such as material
living conditions, education levels, and the like) the two types of missing
values should not make much difference, unless refusers did also refuse to
answer other questions.
Hector
-----Original Message-----
From: SPSSX(r) Discussion [mailto:
[hidden email]] On Behalf Of
Graham Reid
Sent: 16 November 2007 10:23
To:
[hidden email]
Subject: expectation maximization & categorical variables
I need to impute family income for about 1/4 of my sample
(350/1400).
Income is reported in 12 categories. I have 2 questions.
First, can SPSS Missing values/expectation maximization handle a
categorical outcome? I did run the procedure & it is treating it
like a
continous variable. Any suggestions on alternatives?
Second, I have 2 types of missing - Don't know (7% of sample) vs
refused =
18% of sample. Any suggestions on how to handle these differently?
Thanks
Graham Reid
=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text
except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD
=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD
Locked
