# st: RE: A follow-up question on the IVPROB/IVTOBIT.

 From "Nick Cox" To Subject st: RE: A follow-up question on the IVPROB/IVTOBIT. Date Wed, 31 Jul 2002 13:23:23 +0100

```Kee-Lee Lee

> I do not know how to thank you guys for your kind
> helps on my question (on IVPROB/IVTOBIT): with your
> kind help, installing the ado files worked great.
>
> But, if you don't mind, I would really like to ask you
> my follow-up question.
>
> I have been trying to implement the IVPROB procedure
> with the following two equations:
> Eq1) A1 = A2 + A3 + A4 + A5
> Eq2) A2 = A1 + A3 + A6 + A7
>
> Variable A1 is dichotomous and A2 is continuous: some
> of A3-A7 are dichotomous and some of A3-A7 are
> continuous.
>
> I have run the following STATA command to implement
> the IVPROB procedure: ivprob A1, endog(A2) iv(A6 A7)
> exog(A3 A4 A5). And, I got the error message saying
> that "matrix has missing values".
>
> What does that mean, why did it happen, and how can I
> solve it? Is it something about the missing values in
> my variables (I have some missing values on A2 and
> A3-A7, but not A1) or is it something about the
> multicollinearity problem or is it about something
> else.
>

-ivprob- was written by Joe Harkness, so he is the expert.

to the problem you have; even if this is not true,
they could lead to other difficulties.

With the call

ivprob A1, endog(A2) iv(A6 A7) exog(A3 A4 A5)

-ivprob- will automatically ignore any observations which have
missing values on A1, but you have none.

However, different bits of the calculations within -ivprob-,
based on calls to -probit- and -regress-,
will very possibly omit different parts of the
observations given the missing values on A2-A7.

My guess is that if this happens the results will be misleading at
best.

You can test this for yourself. Make sure that the command
ignores all observations with any missing values:

egen nmiss = rmiss(A2 A3 A4 A5 A6 A7)
ivprob A1 if nmiss == 0, endog(A2) iv(A6 A7) exog(A3 A4 A5)

If your problem persists, you need econometricians' help.

Open suggestion on -ivprob-: should not

marksample touse

be followed by

markout touse `endog' `iv' `exog'

Nick
n.j.cox@durham.ac.uk

*
*   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/
```