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

Re: st: Fractional response data and effect sizes

From   Maarten buis <>
Subject   Re: st: Fractional response data and effect sizes
Date   Thu, 27 Nov 2008 14:55:31 +0000 (GMT)

--- "[ISO-8859-1] Natália Barbosa" <> wrote:
> (1) a QMLE Poisson estimator using Stata's glm command with link(log)
> exp(total number of firms) and robust
> (2) the WLS estimator applied to the arcsine-root transformation of
> the proportion of innovative firms
> (3) the fractional logit model proposed by Papke and Wooldridge
> (1996) using Stata's glm command with link(logit) family (binomial)
> and robust

Just a language remark: Models (1) and (3) are very very similar, the
only difference is the link function, but by giving very different
sounding discriptions you are suggesting that there are much more
differences between these models than is actually the case.
> A referee asks me for the effect sizes. How can I measure "effect
> sizes" using STATA and the above alternative procedures?

After -glm- (i.e. for models (1) and (3)) you can use -mfx-. I don't
know how you implemented model (2) but that sounds harder as you seem
to be modeling the average of the transformed proportion rather than
the average of the proportion.

-- Maarten

Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands

visiting address:
Buitenveldertselaan 3 (Metropolitan), room N515

+31 20 5986715

*   For searches and help try:

© Copyright 1996–2019 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index