Re: st: MI of a continuous by categorical interaction term

[daniel klein <klein.daniel.81@gmail.com>]

Caroline,

from my understanding the JVA (just another variable) in this case
will lead to just k - 1  other variables, where k is the number of
levels of your categorical variable, to be created before the
imputation. This is because the approach creates the interaction (or
other forms of transformations) first, just as they would be created
in the complete case analyses. Here is the two step I would go:

1. Create k - 1 indicator variables for the k levels of your
categorical variable
2. Create k - 1 new variables multiplying each indicator with the
continous variable

Best
Daniel

-- 
Hello everyone,

I am doing multiple imputation and would like to impute an interaction
variable using the "just another variable" approach.  My interaction
is between an unordered categorical variable and a continuous
variable.  I am trying to code this as a single variable so it can be
included as a response variable in the imputation model. How would
this single variable be created? For example, say the  categorical
variable is marital status (divorced, married, single. etc) and the
continuous variable is time. It seems wrong just to multiply marital
status (which is coded as 1, 2,3, etc.) by time in Stata to create the
new variable. Is there another way to create it?

Any help would be much appreciated.

Thank you!
Caroline
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/