st: RE: RE: RE: RE: RE: RE: Accessing to a subvector of vector like

[Nick Cox <n.j.cox@durham.ac.uk>]

Thanks; that is a detailed reply. I am happy to leave it to -margins- experts to have a go. 

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


-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Nakelse, Tebila (AfricaRice)
Sent: 07 February 2012 15:49
To: statalist@hsphsun2.harvard.edu
Subject: st: RE: RE: RE: RE: RE: Accessing to a subvector of vector like

Nick the code you've given is correct and work very well. But it seems it not solve my problem which I know is not clearly expose. 
Below I tried to xplicit 
explicit.

sysuse auto
(1978 Automobile Data)

. gen gpm = 1 / mpg
. qui logit  foreign  weight length turn
. mat bl=e(b)

. predict pz, p
. qui regress gpm  length displacement   pz
. mat breg = e(b)
. mat li breg
breg[1,4]
          length  displacement            pz         _cons
y1     .00028888       .000039    -.00406527    -.01058354

. mat li bl

bl[1,4]
       foreign:    foreign:    foreign:    foreign:
        weight      length        turn       _cons
y1  -.00225393   .05080304  -.43106583   12.501797

predictnl Ey =  breg[1,1]*length + breg[1,2]*displacement + breg[1,3]* normal(bl[1,1]*weight+bl[1,2]* length+bl[1,3]* turn+ bl[1,4]) +  breg[1,4],  se(Ey_se)
Warning: prediction doesn't vary with respect to e(b).

I want to use Ey in my margins command

margins, expression (breg[1,1]*length + breg[1,2]*displacement + breg[1,3]* normal(bl[1,1]*weight+bl[1,2]* length+bl[1,3]* turn+ bl[1,4]) +  breg[1,4])
                                                                                                                                                                                                                          ||   
                                                                                                                                                                                                                         pz
------------------------------------------------------------------------------
             |            Delta-method
             |     Margin   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
       _cons |   .0501869          .        .       .            .           .
------------------------------------------------------------------------------

The problem here is that, margins does not compute standard error. May be, because the element of  breg and bl are not recognized as element of e(b) . I really do not know why ? 
May be if there is a solution to tell to margin that this elements are from estimation it would solve the problem.

What I have done is to stack bl and breg in one matrix and ereturn  repost it  in  e(b) matrix  used by margins.

But here the main problem is that pz(predicted probability)  is estimated in a first step  using a different set of independent variables.  May be if I extract the coefficients of pz from e(b), margins would take into account the first step . 

And I do not  how to tell margins to take into account the first step.

                                                                                                                                                                        
Best regards 

Tebila 

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Nick Cox
Sent: Monday, February 06, 2012 8:07 PM
To: 'statalist@hsphsun2.harvard.edu'
Subject: st: RE: RE: RE: RE: Accessing to a subvector of vector like

I think you've changed the question. 

Please show us the exact code you used and explain in what sense it is "not working". Otherwise my answer can only be a guess that you did something wrong. 

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

Nakelse, Tebila (AfricaRice)

Thank you Nick 
I tried this solution.  But it still not working. 
I want to use expression(pnl_exp) in margins. But if I use this solution stata may not know whether this coefficients are from a estimation and may not  use this information to compute the matrix of covariance for example.

Nakelse, Tebila (AfricaRice)

Thank you Nick. It is okay.

Nick Cox

Just copy to a regular vector first: 

sysuse auto
gen gpm = 1 / mpg
regress gpm  weight length displacement

. mat li e(b)

e(b)[1,4]
          weight        length  displacement         _cons
y1     .00001174     .00006458     5.517e-06     .00152916

. mat b = e(b)

. mat b1 = b[1, "weight".. "length"]

. mat li b1

b1[1,2]
       weight     length
y1  .00001174  .00006458


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

Nakelse, Tebila (AfricaRice)

I would like to access to a subvector of a vector like e(b) or e(V). 
Stata allows to have et sub-element as _b[varname]. Is it possible to have _b[varlist] ?


*
*   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/

*
*   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/