Syed,
You may get more-useful suggestions if you provide more information
about your data (including the sample sizes) and, especially, about
the parameterization that you are using for the lognormal
distribution.
As Gordon Hughes pointed out, the usual lognormal distribution covers
only nonnegative values and has two parameters (sometimes defined as
the mean and variance of the corresponding normal distribution in the
log scale). Often a third parameter is added to shift the origin away
from zero (usually to some positive value). Your earlier posting
related "shift" to skewness, but the usual shift parameter has no
effect on skewness. Sometimes data come from a truncated distribution
(e.g., data below a specified threshold are not reported). That
threshold or truncation point may not be the same as the shift. Also,
the choices in analysis depend on whether one knows the number of
observations whose values were below the truncation point.
Depending on those features of the data, I might consider an empirical
quantile-quantile plot of the two distributions before getting into
any formal inferences.
David Hoaglin
On Sat, Jan 5, 2013 at 10:55 PM, Syed Hasan <mhasan26@yahoo.com> wrote:
> Thanks Nick. I appreciate your response.I should have been clear on my question. My intent is to compare truncation of one distribution compared to the other. Sorry about that.
>
> Regards,
>
> Syed
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/faqs/resources/statalist-faq/
* http://www.ats.ucla.edu/stat/stata/
