Thanks Nick,
the more I think about this the more I tend to agree that log(variable +
fudge) is no solution to my transformation problem.
Using -glm- is a nice idea. Unfortunately, inflation is not my response
variable. The goal of the transformation of X (inflation rates) is to
calculate various conditional means of X|Y (e.g. mean inflation rate in
region Z in year C; or mean inflation rates of countries with property B in
a given year etc.) and use these as IV in my logit model.
The calculations of these means should not be too heavily influenced by the
outliers (I do not want to use medians, since the outliers should enter the
calculations of the averages).
Best,
Jens
> I don't think there is an easy quick fix
> in this situation. > in my view creates as many problems as it
> solves.
> What sometimes help is to use -glm- with
> say -link(log)-. The crucial detail is that
> -glm- does _not_ depend on log(observed)
> being determinate. That is relevant if inflation
> rate is your response variable.
> Nick
> [email protected]
[email protected]
> this one is probably a softball for most of you - but is
> there any smart
> mathematical transformation in STATA that allows me to deal with the
> problem of outliers while keeping the distances between the
> values of my
> variable in proportion.
>
> The variable I would like to transform contains annual inflation rates
> ranging from -30.2 to 23773 so taking LNs does not work because of the
> negative values. I thought about simply adding 30.2 to each
> value and then
> taking the LNs, but I am not quite sure if this is a legitimate way of
> dealing with my problem.
*
* 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/
*
* 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/
