Re: st: ice error

[A Loumiotis <antonis.loumiotis@gmail.com>]

Dear Fernando,

-ice- (SJ 9-3) is a user written program.  Please remember that you
are asked to explain where user written programs come from.

My guess is that the error is caused because of the stepwise option.
You may try to run -ice- without it.  Alternatively, you may also need
to specify option swopts(group(group_list)) where the group_list
includes your ordered categorical variables (you might need to specify
them using an i. prefix instead of an o. prefix).  Also make sure that
none of the variables that nbreg will be used are included in the
mainvarlist with an o. prefix.  These are my guesses about what
perhaps is going wrong.

Best regards,
Antonis

On Mon, Jan 23, 2012 at 9:42 PM, fernando andrade
<fandrade.stat@gmail.com> wrote:
>
> Dear all
> i am using ICE to perform multiple imputation in a large set of
> variables, while i am running the dry run i found myself with a
> problem i do not expected.
> ice is giving me this error message for several ordinal variables that
> were speficied with the o.varname option:
>
> Error! _Iy1rel5_w2_2 found on both sides of prediction equation
> Error! _Iy1rel5_w2_3 found on both sides of prediction equation
> Error! _Iy1rel5_w2_4 found on both sides of prediction equation
> Error! _Iy1rel5_w2_5 found on both sides of prediction equation
>
> in this case the variable is y1rel5_w2 and the ice command i am using is
>
> ice a1alc8a_m   debt_w2 y1cig   y  ...  ///
> m.y1duom5c_w2 m.y1duom5d_w2 ...  m.t1050 ///
> o.y1sc101  o.y1rel5_w2 ....     o.y1pp1k_w2     ///
> y1nei5  ....    t1517 ///
> conditional(a1htv1 : flagadult==1 \ a1htv2 : flagadult==1 ...
> \       T1016as116 : flagactua ==1 ) ///
> cmd(a1pm15a  ...  y1alc16b_w2: nbreg) ///
> seed(123) ///
> stepwise ///
> m(10) ///
> dryrun
>
> note: "..." in the command mean that i omitted var names and other to
> make the command easy reading
>
> I understand that ICE is generating dummies for  each variable
> specified as o.varname or m.varname, (as in the case of o.y1rel5_w2)
> and uses this dummies as predictors, but ice should not try to
> directly impute the set of dummies that generates based on the
> o.varname specification, right?
>
> I appreciate if you could give some ideas as why i get these errors
>
> thank you
>
> Fernando
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/