RE: st: FW: Elasticity using nlcom

[HENRY AKAEZE <akaezehenry@yahoo.com>]

Martin,

You are right: it doesn't make sense to multiply by zero, but stata was able to choose the right value for "a" and "b". When I reset the global to no-zero it still gives me the same value for a and b. It also gave the same values when I did not set a and b to any constant. Please stata output



program nlAI
      version 8.0
       if "`1'"=="?"{
        global S_1"g0 b a d2 y"
        global g0=0.1
        global d2=0.1
       global y=0.1
        exit
      }
    replace `1'=$g0+$b*(lagimportca+$a* laglocalca)-$a*predictlocal+( (gdpca-$b* cifprice1*( lagimportca+$a* laglocalca))/ cifprice1)*($g0-($y+$d2)*ln( cifprice1)+$d2*ln(cpi)+$y*ln(( gdpca-$b* cifprice1*(lagimportca+$a* laglocalca))/priceindex1))
 end
nl AI importca
(obs = 46)

Iteration 0:  residual SS =  520270.6
Iteration 5:  residual SS =  138.2019

      Source |       SS       df       MS            Number of obs =        46
-------------+------------------------------         F(  5,    41) =     71.68
       Model |  1208.15295     5   241.63059         Prob > F      =    0.0000
    Residual |  138.201885    41  3.37077768         R-squared     =    0.8974
-------------+------------------------------         Adj R-squared =    0.8848
       Total |  1346.35484    46  29.2685834         Root MSE      =  1.835968
                                                     Res. dev.     =  181.1458
(AI12)
------------------------------------------------------------------------------
    importca |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
          g0 |  -.0124388   .0305046    -0.41   0.686     -.074044    .0491665
           b |   .7468116   .1050251     7.11   0.000     .5347091    .9589141
           a |   .1386389   .0638289     2.17   0.036     .0097338     .267544
          d2 |   .0011694   .0031635     0.37   0.714    -.0052194    .0075581
           y |   .0019213   .0048115     0.40   0.692    -.0077958    .0116384
------------------------------------------------------------------------------
 (SEs, P values, CIs, and correlations are asymptotic approximations)



nlcom ((-.0093/cpi)*((gdpca/cifprice1)-$b*(lagimportca-$a*laglocalca))*8.89)


--- On Mon, 6/7/10, Martin Weiss <martin.weiss1@gmx.de> wrote:

> From: Martin Weiss <martin.weiss1@gmx.de>
> Subject: RE: st: FW: Elasticity using nlcom
> To: statalist@hsphsun2.harvard.edu
> Date: Monday, June 7, 2010, 2:38 PM
> 
> <>
> 
> This only deepens the mystery for me, to be honest. Why
> would you want to
> multiply -(lagimportca-$a*laglocalca)- by zero? If you
> really want "...point
> elasticity estimates, standard errors, test statistics,
> significance levels,
> and confidence intervals for nonlinear combinations of
> parameters", why do
> you not put the parameters into your -nlcom- call, but
> constants?
> 
> 
> HTH
> Martin
> 
> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu
> [mailto:owner-statalist@hsphsun2.harvard.edu]
> On Behalf Of HENRY AKAEZE
> Sent: Montag, 7. Juni 2010 22:57
> To: statalist@hsphsun2.harvard.edu
> Subject: Re: st: FW: Elasticity using nlcom
> 
> Hello Martin,
> 
> Thanks for the quick reply. My globals for "a" and "b" are
> zeros (0).
> Indeed, I intend to compute point elasticity estimates,
> standard errors,
> test statistics, significance levels, and confidence
> intervals for nonlinear
> combinations of parameters. The nlcom is post estimation
> command after using
> writing and estimating an nl program in stata.
> 
> Thank you.
> 
> Henry
> 
> --- On Mon, 6/7/10, Martin Weiss <martin.weiss1@gmx.de>
> wrote:
> 
> > From: Martin Weiss <martin.weiss1@gmx.de>
> > Subject: st: FW: Elasticity using nlcom
> > To: statalist@hsphsun2.harvard.edu
> > Date: Monday, June 7, 2010, 12:38 PM
> > 
> > <>
> > 
> > So Henry sent this query to me privately. I am not
> sure
> > what his -global-s
> > "a" and "b" contain, but from his second example, I
> think
> > they represent
> > scalars. Henry should think about the point of using
> > -nlcom-: "nlcom
> > computes point estimates, standard errors, test
> statistics,
> > significance
> > levels, and confidence intervals for (possibly)
> nonlinear
> > combinations of
> > parameter estimates after any Stata estimation
> command."
> > (see the help file)
> > 
> > So when Henry feeds -nlcom- his call
> >
> -nlcom((-.0093/cpi)*((gdpca/cifprice1)-$b*(lagimportca-$a*laglocalca))*8.89)
> > -, Stata sees variable names instead of parameter
> > estimates, and the error
> > message makes the problem abundantly clear. Let me
> > replicate:
> > 
> > **
> > sysuse auto, clear
> > reg price weight length
> > nlcom (myratio: weight/length )
> > **
> > 
> > His second -nlcom- call is:
> > -nlcom((-.0093/35)*((36/.33)-$b*(3.7-$a*14.4))*8.89)-
> which
> > Stata evaluates
> > to -nlcom
> >
> ((-.0093/35)*((36/.33)-.6892200041267298*(3.7-.1785226557738297*14.4))*8.89)
> > -. This, of course, does not contain a reference to a
> > parameter estimate,
> > either, so Stata returns a scalar, and no other
> values.
> > Where is a standard
> > error supposed to come from for a _scalar_? 
> > 
> > All of which makes me think that Henry really wants
> > something else, and my
> > money is on -predictnl-.
> > 
> > 
> > HTH
> > Martin
> > 
> > -----Original Message-----
> > From: HENRY AKAEZE [mailto:akaezehenry@yahoo.com]
> > 
> > Sent: Montag, 7. Juni 2010 20:46
> > To: martin.weiss1@gmx.de
> > Subject: Elasticity using nlcom
> > 
> > Dear Martin,
> > 
> > I am doing estimation using the AIDS model. In order
> to
> > estimate elasticity,
> > I used nlcom command on an elasticity function I
> derived
> > through
> > differentiation as follows:
> > 
> >
> nlcom((-.0093/cpi)*((gdpca/cifprice1)-$b*(lagimportca-$a*laglocalca))*8.89)
> > 
> > The exogenous variables are CPI, CIF and the lags of
> import
> > and domestic
> > products.
> > 
> > However, stata complained that reference was made to
> X
> > rather
> > than _b[X] as follows:
> > 
> > 
> > 
> >
> expression((-.0093/cpi)*((gdpca/cifprice1)-.6892200041267298*(lagimportca-.1
> > 785226557738297*laglocalca))*8.89)
> > contains  reference to X rather than _b[X]
> > 
> > When evaluated at their means I got a value but
> without
> > standard errors and
> > p-value as below:
> > 
> >  
> nlcom((-.0093/35)*((36/.33)-$b*(3.7-$a*14.4))*8.89)
> > 
> >  
> >
> _nl_1:(-.0093/35)*((36/.33)-.6892200041267298*(3.7-.1785226557738297*14.4))*
> > 8.89
> > 
> > 
> > 
> >
> ----------------------------------------------------------------------------
> > --
> > 
> >     importca
> > |             
> > P>|t|      Coef.     
> >                
> >         [95% Conf. Interval]
> >                
> >        Std. Err.   
> >   t
> >
> -------------+--------------------------------------------------------------
> > --
> > _nl_1 -.255856          . 
> >             .   
> >            
> >    .     
> >                
> >      
> > 
> >
> ----------------------------------------------------------------------------
> > --
> > 
> > I need to know what is going on and what I must do to
> get
> > values for the
> > standard errors,  t-stat and conf. interval.
> > 
> > Thank you,
> > 
> > Henry
> > 
> > 
> >       
> > 
> > *
> > *   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/
> 


      

*
*   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/