Hi,
I have panel data collected in 3 waves and want to do Multiple
Imputation by Chained Equations using the command <<ice>>. Let's say
the dependent variable at time 1 is <y1>, and the predictors <x1> and
<z1>. <x1> is a multinomial variable with three categories: <x1a>,
<x1b> and <x1c>. Since I have repeated measures for each variable, I
don't know how to specify the equations. I also want to verify that I
am handling the multinomial variable in the correct way. Here is what
I did:
<<ice y1 y2 y3 x1 x2 x3 x1a x2a x3a x1b x2b x3b x1c x2c x3c z1 z2 z3
using imputed_database.dta, replace m(5)
passive(x1a:x1==1\x1b:x1==2\x1c:x1==3 \ x2a:x2==1\x2b:x2==2\x2c:x2==3
\ x3a:x3==1\x3b:x3==2\x3c:x1==3) substitute(x1: x1a x1b x1c, x2: x2a
x2b x2c, x3: x3a x3b x3c) eq(y1: y2 y3 x2a x3a x2b x3b x2c x3c z1 z2
z3, y2: y1 y3 x1 x2 x3 x2a x3a x2b x3b x2c x3c z1 z2 z3 , y3: y1 y2 x1
x2 x3 x2a x3a x2b x3b x2c x3c z1 z2 z3, x1: y1 y2 y3 x2 x3 x2a x3a x2b
x3b x2c x3c z1 z2 z3, x2: y1 y2 y3 x1 x3 x2a x3a x2b x3b x2c x3c z1 z2
z3, x3: y1 y2 y3 x1 x2 x2a x3a x2b x3b x2c x3c z1 z2 z3, z1: y1 y2 y3
x1 x2 x3 x2a x3a x2b x3b x2c x3c z2 z3, z2: y1 y2 y3 x1 x2 x3 x2a x3a
x2b x3b x2c x3c z1 z3, z3: y1 y2 y3 x1 x2 x3 x2a x3a x2b x3b x2c x3c
z1 z2)>>
As you can see, for each equation I excluded from the prediction
equation: (1) the variable being imputed, (2) the multinomial variable
<x>, and (3) the dummy variable <xa>. I wonder if I should have also
excluded predictors measured in a later wave than the variable being
imputed (for example, dropping all the variables measured in wave 3
when imputing <y2>).
I would appreciate your feedback regarding the way I am using (or
should be using) the <<ice>> command.
Thanks,
Esteban.
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
