Small challenge for experienced SPSS syntax programmers

[Jo Blois]

Hi, I am dealing with a database with missing values for some
variables.
In order no to looses cases in my regression models, I would like to
generate values for the missing ones that correspond to the mean value
of the specific variable. I will test afterwards if the new models
(with the created values) acts in a similar fashion than the one with
the missing values.

Here is the question. Instead of generating a value equal to the mean
for each missing value, how can we make a random generation of values
that have the same mean but normally distributes within the range of
two SD.

As a concrete example. One continuous variable can be from 1 to 100.
The mean is 33. The SD is 12. There are 4000 cases and 20 % missing
values.

So insteand of attributing the value of 33 for all missing values, I
would like that the new created values for the 20 % missing have a
mean of 33 but would be randomly assigned and would range from 21 to
45.

Anyone is capable of doing so?

[Bruce Weaver]

If you have v17 or later, why not use MULTIPLE IMPUTATION?

--
Bruce Weaver
bwe...@lakeheadu.ca
http://sites.google.com/a/lakeheadu.ca/bweaver/Home
"When all else fails, RTFM."

[Jo Blois]

I am afraid this is a add-on that I do not have access to.

[Rich Ulrich]

This sounds like a very bad idea.

Here is how to do it.

First, replace the Missing's with the mean, and create
an indicator variable (0/1 for No/Yes Missing) - just like
you would use if you wanted to include the variable and
its indicator in an ordinary regression.

Then generate your truncated normal deviation-scores--
mean= 33; SD= 12; tests to truncate or drop extremes;
add the Deviation-score to the variable.

--
Rich Ulrich