Would anyone on the list be kind enough to give me advice on how
to implement the procedure for estimation of variance for non-
linear combinations of parameter estimates (e.g. elasticities),
as described by Krinsky & Robb (1986), in Stata?
I use data from a Discrete Choice Expmeriment to estimate a
choice model with xtlogit, i.e.
-xtlogit choice H1 H2 H3_varm H3_kold H4 H5, re-
where H* are the attributes describing the choice situation.
Afterwards, I calculate Willingness-to-Pay measures, e.g.
(WTP_H4: - _b[H4]/_b[H5])
where H5 is the price attribute.
I would like to estimate the variance of WTP using the
(simulation) method proposed by Krinsky-Robb or alternatively
another simulation method (for example, I know how to bootstrap
the original parameter estimates, but is it possible to obtain
bootstrap estimates of var(WTP)?).
I know that I can obtain a very neat output with WTP's and
variance estimates, using -nlcom-, which uses the delta method
for estimating variance estimates. I have already done that. I
would like to compare the various variance estimates obtained,
which is why I'm trying to implement alternative methods.
-- J. K. Hansen
For an example, see:
view http://www.nat.sdu.dk/users/nat-sdu/jkha01/data/jorgen.txt
or do:
do http://www.nat.sdu.dk/users/nat-sdu/jkha01/data/jorgen.do
Reference:
Krinsky, Itzhak & A. Leslie Robb (1986), "On Approximating the
Statistical Properties of Elasticities", The Review of Economics
and Statistics, MIT Press, vol. 68(4), pages 715-19, November.
*
* 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/