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 at the end of May, and its replacement, statalist.org is already up and running.


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

Re: st: update betafit available from SSC


From   Nick Cox <njcoxstata@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: update betafit available from SSC
Date   Fri, 8 Apr 2011 16:29:18 +0100

This version is beta in every way!

A favourite personal detail: the help file for -betafit- links to a
book by MacKay with a very neat section showing that the logit of a
beta-distributed variable has itself a bell-shaped distribution --
even if that beta distribution is U-shaped. The text of that book is
publicly available on the internet. So, although this is _not_ logit
modelling, beta distributions mesh with logit scales very nicely.
(Apologies if you know this already or regard it as intuitively
obvious.)

Note to people who think volatility: not your beta.

Nick

On Fri, Apr 8, 2011 at 4:10 PM, Maarten buis <maartenbuis@yahoo.co.uk> wrote:
> Thanks to Kit Baum an update of the
> -betafit- package is now available from SSC. To
> install type in Stata:
> -ssc install betafit, replace- or
> -adoupdate, update-.
>
> -betafit- fits by maximum likelihood a beta-
> distribution with or without covariates to data.
> One of the key characteristics of the the beta-
> distribution is that it is a fairly flexible
> distribution for variables that are bounded
> between 0 and 1. It can thus be used to model
> proportions or other bounded variables, and
> optionally how they depend on explanatory/
> independent/right-hand-side/x variables.
>
> This update contains the following changes:
>
> o After estimating a -betafit- model one can
>   predict 3 new types of residuals: working
>   residuals, partial residuals, and score
>   residuals.
>
> o For users who have Stata 11 or higher: -betafit-
>   allows factor variables and with that also the
>   use of -margins- as a post-estimation tool.
>
> o -betafit- can display results as "relative
>   proportion ratios", that is, as exponentiated
>   coefficients. A discussion on relative
>   proportion ratios is included in the help file.
>
> -betafit- has a relatively long history with input
> from various authors.
>
> In 1997 Nick Cox wrote a -beta- for fitting a beta
> distribution by ML without covariates. This program
> is still part of -betafit- as -beta4-, and it
> should work with Stata 4 upwards. This didn't use
> -ml-: it was a fairly straight port of an algorithm
> published in Journal of Applied Meteorology by
> P.W. Mielke in 1975.
>
> In 2003, riding on the back of some work by Stephen
> Jenkins on other distributions, Nick Cox wrote a new
> -betafit- that now took covariates if specified. But
> it stuck to the parameterisation of a beta
> distribution in terms of two shape parameters.
>
> Every now and again threads on Statalist ask what
> should be done when the response is a continuous
> proportion between 0 and 1, and sooner or later
> someone mentions the beta distribution as one
> possibility, but that that the then existing
> parameterisation of -betafit- was not particularly
> useful for problems with covariates.
>
> In 2005 I came on board, taking the initiative
> in implementing an extra parameterisation in terms
> of a location and a scale parameter. Since then I
> have initiated various extensions to the program
> like adding capabilities to compute marginal effects
> (2006), allow the use of -predict- (2008), and the
> latest update that became available today. -betafit-
> is now attributable to Buis, Cox, and Jenkins.
>
> I hope that some of you will find it useful,

*
*   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