Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: Xaxis transformation after logging variable


From   Nick Cox <njcoxstata@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Xaxis transformation after logging variable
Date   Tue, 21 Aug 2012 17:00:05 +0100

Of your three references here, one is now recognisable to me as a
paper in Biometrika. The others ring no bells without details. The
parameterisation details make little sense without any definitions or
formulas.

It's possible that this distribution is a reparameterisation of one
that can be fitted with code by Stephen Jenkins with/without
coauthors. I suppress full references to underline what you are (not)
doing here.

Otherwise, I advise reposting.

Before you do that, please do read the Statalist FAQ.

Nick

On Tue, Aug 21, 2012 at 4:24 PM, Francisco Augusto
<francisco.augusto.7@gmail.com> wrote:
> Thanks for the quick answer. It worked! (I am sorry for the trivial
> question, though)
>
> The reference I was talking about was Nick Cox, Speaking Stata:
> Graphing Distributions, 2004, The Stata Journal, pp 77
>
> For the other question, I am sorry for the lack of information.
>
> I am trying to use the Prentice 1974 "A log gamma model and its
> maximum likelihood estimation" Biometrika Trust. In this approach the
> parameter k is transformed to q =  k ^ (-1/2) with support to k of 0
> to +infinity , parametrization also used by Lawless 1980 "Inference in
> the generalized gamma and log gamma Distributions" and applied in
> Cabral & Mata 2003 "On evolution of firm size distribution: facts and
> theory".
>
> With this approach I intended to apply the extended generalized gamma
> distribution to the logged variable, and by that obtaining estimates
> for the different parameters. After that, I was trying to regress the
> logged variable on other variables assuming the logged variable
> follows the extended generalized distribution.
>
> Should I "repost" the previous question? Is this information enough?
>
> Thank you very much for all the help,
> Francisco Augusto
>
>
>
>
> On Tue, Aug 21, 2012 at 3:50 PM, Nick Cox <njcoxstata@gmail.com> wrote:
>> "Cox showed" (even guessing and narrowing it down to me) is not a
>> precise reference. But if your axis variable is log_10(variable) and
>> you want labels in terms of variable, you can just go something like
>>
>> ... xla(0 "1" 1 "10" 2 "100")
>>
>> This is well documented at -help axis label options- and is no sense
>> limited to showing kernel density estimates.
>>
>> I guess people ignored your previous question because it wasn't clear
>> (there are several possible generalisations oof the gamma
>> distribution) and was based on an incomplete reference. Still true.
>>
>> Nick
>>
>> On Tue, Aug 21, 2012 at 3:27 PM, Francisco Augusto
>> <francisco.augusto.7@gmail.com> wrote:
>>> Dear Statalist,
>>>
>>> I am dealing with a small problem: I am doing kernel (kdensity) of a
>>> logged variable and I would like to rearrange the Xaxis to be in the
>>> original scale and on a 10 exponential order (like 1 10 100 1000
>>> 10000, of the original values). I have seen a solution for the first
>>> part of the problem: Cox showed a command to get the original scale on
>>> the xaxis after logging a variable. Nevertheless, I don't know how to
>>> have the  1 10 100 1000 10000 order.
>>>
>>> Plus, I am still struggling to find a solution for the Prentice 1974
>>> generalized gamma approach. It would be so nice for someone to present
>>> a possible solution!
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index