Thank you Ekrem,
In fact, your solution works but it is not exactly what I am looking for.
Here is a simplified example to be more precise. Consider I have only
two explanatory variables, a constant and time t, so that my parameters
‘a’ and ‘b’ are given by:
a = a0 + a1*t
b = b0 + b1*t
what I get from -nl- looks like:
------------------------------------------------------------------------------
| Robust
CumPcHa | Coef. Std. Err. t P>|t| [95% Conf.
Interval]
-------------+----------------------------------------------------------------
/a0 | .642167 .0098387 65.27 0.000 .6228815
.6614525
/a1 | .0011107 .0003833 2.90 0.004 .0003593
.0018621
/b0 | .5155133 .0091666 56.24 0.000 .4975453
.5334813
/b1 | -.0017981 .0003266 -5.51 0.000 -.0024383
-.0011579
------------------------------------------------------------------------------
now, if I run -nlcom- with a particular value of t to compute my Gini
coefficient ‘g’, the result is something like:
------------------------------------------------------------------------------
CumPcHa | Coef. Std. Err. t P>|t| [95% Conf.
Interval]
-------------+----------------------------------------------------------------
g | .5759562 .0022265 258.68 0.000 .5715919
.5803205
------------------------------------------------------------------------------
ok, I have an estimate of ‘g’, its standard error, t and P statistics
and confidence interval but what I really would like to have are the /g0
and /g1 (equivalent to /a0, /a1, /b0 and /b1) that is the coefficients
of ‘g’ associated with the constant and time variables (or in other
words the impact of time on ‘g’).
Any clue ?
Thank you in advance,
Laurent
Ekrem Kalkan a écrit :
Hello,
you need to take a particular point of X , for example its mean or
minimum or maximum value.
You can do this:
sum X
gen xbar=r(mean)
and then insert "xbar" where it is needed in your nlcom formula
I hope this is what you asked.
Ekrem Kalkan
2009/12/16 Laurent Piet <[email protected]>:
Dear list,
I think my problem can be solved with -nlcom- but I have not figured out how
to do yet.
I use the “function evaluator program” version of the -nl- command to fit a
parametric Lorenz curve to observed data. The two parameters ‘a’ and ‘b’
characterizing the functional form of the Lorenz curve are expressed as
linear combinations of explanatory variables, say a = a0 + a1*X1 + a2*X2 and
b = b0 + b1*X1 + b2*X2. So what I really get from -nl- are the a`i's and
b`i's coefficients along with their associated std. err., t and p
statistics.
Now, I want to compute a third parameter, the Gini coefficient ‘g’, which is
a non-linear combination of ‘a’ and ‘b’ involving the Beta function. What I
ultimately want to compute are the g`i's, that is, the coefficients
measuring the effects of the X`i's (including the constant) on ‘g’, with
their associated std. err., t and p.
I know how to use -nlcom- to do so when only the constant term is involved
(nlcom g0: f(_b[/a0],_b[/b0])), but I couldn’t find how to deal with the
explanatory variables X.
Any idea would be welcome !
Thanks in advance,
Regards,
Laurent
--
Laurent PIET
----------
INRA, UMR1302 INRA-Agrocampus
"Structures et Marchés Agricoles, Ressources, Territoires"
4 allée Adolphe Bobierre | tél. : +33 (0)2.23.48.53.83
CS 61103 | fax. : +33 (0)2.23.48.53.80
F-35011 Rennes cedex | [email protected]
----------
http://www.rennes.inra.fr/smart
*
* 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/
