[Limdep Nlogit List] Statistical significance of demand elasticities

avassilopoulos@aua avassilopoulos.aua at gmail.com
Thu Jan 14 04:50:45 EST 2010


Dear all,

 

I'm utilizing Shonkwiller and Yen's variation of Heckman's two-step method
to derive demand elasticities for several product categories, for which I
would also like to examine their statistical significance. 

Since the elasticities are a function of both coefficients from the first
stage (probit) and the second stage (OLS), wald test is not applicable and
bootstrapping the whole procedure seems to be the only solution(?) to my
problem. 

On the other hand, my procedure includes two estimations, Murphy and Topel's
variance-covariance matrix correction, several matrix and scalars
calculations (needed for all the above) and of course the calculations of
elasticities; so it far exceeds the limitations of limdep procedures, both
in terms of nonblank characters and in terms of commands' number. I've
already tried to brake down the commands in small procedures and run them
all in a final one but limdep doesn't recognize procedure-in-procedure
commands. Moreover, in the past limdep crashed after the second or third
simulation, every time I had bootstrapped a relatively small procedure but
with several MATRIX; CALC; CREATE; and KEEP; commands.

 So my question is twofold: Is there a way to bootstrap very large and
complicated procedures in limdep? If not, is anyone aware of any other way
to examine the statistical significance of a variable (demand elasticity in
my case) which is the product of coefficients derived from more than one
estimations, and thus with unknown covariance matrix?

 

Thanking you in advance,

------------------------- - ------------------------------

Achilleas Vassilopoulos

                                        

 



More information about the Limdep mailing list