Zou Hong <[email protected]>:
Your problem seems ideal for -poisson- (or -glm- or -zip-); see
http://www.statacorp.com/statalist/archive/2007-04/msg00549.html
and related posts for some relevant discussion.
Suppose "ins" is insurance and "asset" is value of insurable
assets--then you should
. g lna=ln(asset)
. poisson ins lna size1 tang, r
to get an estimate of the elasticity of insurance with respect to
assets. Your current model *assumes* the elasticity is one, i.e. the
model with the coef on lna constrained to one is equivalent to
. g ratio=ins/asset
. poisson ratio size1 tang, r
but you do not know that the coef is equal to one, right?
As an example, consider:
webuse nlswork, clear
keep if year>78
bys id (year): drop if _n<_N
gen total=wks_work*hours
gen wage=exp(ln_w)
gen earn=wage*total
gen lntot=ln(total)
poisson earn lntot grade tenu, r
test lntot=1
poisson wage grade tenu, r
Note that just because you can't reject the null hypothesis that the
coef on the "denominator" variable is one, doesn't mean you should
blindly accept the null. (There is a positive association between
higher hours and higher wages in the example, and there likely is a
similar pattern in your insurance and assets data.) It just makes
sense to estimate the parameter IMHO.
On 7/22/07, Zou Hong <[email protected]> wrote:
Dear lister,
I am investigating the insurance consumption issue using Cragg's (1971)
model, which is a first-stage probit plus a second-stage truncated model at
zero (using only firms buying insurance). My sample is quite big around
60,000 observations rougly with 50% of firms buying insurance.
I then find "ins1", the dependent variable defined as insurance
expense/total assets, is higly skewed with a high kurtosis (see descriptive
statistics below). I suspect this is the source of the problem. To mitigate
the skewness, I create a variable "lnins1" (= ln (1+ins1*1000)) that is
truncated at 0. I multiply ins1 by 1000 since ins1 is a very small ratio
variable. I then reestimated the truncated variable and the model did
converge (see below).
I wonder whether my above transformation makes a sense. I think it does
preserve the interpretation of the direction of independent variables on
"ins1".
Any and suggestions comments are welcome
Joe
*
* 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/
