Hi,
If I remember correctly, when you run a negative binomial model in stata, it
includes some sort of hypothesis test to determine whether or not there is
enough overdispersion to conclude that the negative binomial model is an
appropriate alternative to the poisson. Is it possible that this is what
you're seeing? If you have access to stata, you might try running the
regression (the command is "nbreg") to see if it looks like that's what
the
author is reporting.
Good luck,
Dennis
-----Original Message-----
From: gov2001-l-bounces(a)lists.fas.harvard.edu
[mailto:gov2001-l-bounces@lists.fas.harvard.edu] On Behalf Of Abby
Williamson
Sent: Thursday, March 30, 2006 6:31 PM
To: gov2001-l(a)lists.fas.harvard.edu
Subject: [gov2001-l] Alpha (Heterogeneity Coef.)
Hello All,
After extensive sleuthing, we still have a mystery coefficient in one of the
less crucial tables in our paper.
The table includes three models of a negative binomial regression with the
size of discussion networks as the independent variable. The explanatory
variables are (1) whether the respondent came from a 1985 or 2004 wave of
the survey, (2) the interviewers' perceptions of the respondent's
cooperativeness (cooperative, restless, hostile, with friendly/interested as
the omitted base case), (3) the number of the previous 10 questions the
respondent had refused to answer, and (4) demographic characteristics of the
respondents (education, sex, age, marriage status, race). Finally, each
model includes the following coefficient, labeled only as "Alpha
(Heterogeneity Coef.)," followed by an F-test.
We can't find anything in the article on what this might be. It doesn't
seem like it would be appropriate to use a Cronbach's alpha here (and the
coefficients are really low - 0.059 to 0.159, with the values getting lower
as additional explanatory variables of the above added to the model). And
if they are using Blau's heterogeneity index, we can't figure out how.
Does this jog anyone's memory? If not, we'll contact the authors, but I
wanted to try this last ditch effort.
Many thanks,
Abby
-----Original Message-----
From: gov2001-l-bounces(a)lists.fas.harvard.edu
[mailto:gov2001-l-bounces@lists.fas.harvard.edu]On Behalf Of Ian Brett Yohai
Sent: Thursday, March 30, 2006 4:54 PM
To: gov2001-l(a)lists.fas.harvard.edu
Subject: Re: [gov2001-l] Quantities of Interest
Hi Geoff,
Yes, exponential models can fit multiple independent variables - like any
other sysetmatic component.
Best,
Ian
On Thu, 30 Mar 2006 ghumphr(a)fas.harvard.edu wrote:
One last question on this subject (and a few words of
praise for
Zelig, an excellent package)- can exponential models be fit reliably
to multiple independent variables?
I hope that all is well with you.
Quoting ghumphr(a)fas.harvard.edu:
> I took a closer look and decided that the whole thing is garbage.
> Still working on it (spring break and replication pretty much done).
>
> Quoting Ian Brett Yohai <yohai(a)fas.harvard.edu>du>:
>
> > Hi Geoff,
> >
> > Exponential regression models are duration models (time until an
> > event
> > happens) - so I'm not quite sure what you mean when you say you
> > want
to
> > adapt it to time-series analysis. Also
zelig fits exponential models.
> >
> > Best,
> > Ian
> >
> > On Wed, 29 Mar 2006 ghumphr(a)fas.harvard.edu wrote:
> >
> > > Hi,
> > >
> > > I was thinking about how to do exponential regression and came
> > > up
with
> this
> > > quick optimization. I was considering using it to fill in
> > > missing
parts
> of
> > a
> > > Zipf distribution, but I have decided that the assumptions to do
> > > so
are
> not
> > > met, particularly considering that political manifestos in
> > > German
(which
> > has a
> > > lot of compound words) have a pretty well-filled Zipf
> > > distribution
> whereas
> > > manifestos in English (which has a lot of multi-word terms) do not.
I am
> > > curious as to how to adapt this
estimator to time-series
> > > analysis
and to
> > make
> > > it more robust.
> > >
> > > Geoff
> > >
> > >
> > > # the function to optimize
> > > f <- function(par, X, Y, W=rep(1, nrow(X))) {
> > > beta <- par
> > > beta[2] <- 1/par[2]
> > > sum(t(W) %*% abs(exp(X %*% beta) - Y)) }
> > >
> > > # read in the data and set up variables table <-
> > > read.csv("table.csv", header=T) Y <-
as.matrix(cbind(table[7]))
> > > X <- as.matrix(cbind(1, rev(c(1:nrow(Y))))) W <-
> > > as.matrix(table[6])
> > >
> > > # do a linear regression on transformed Y values, taking the #
> > > reciprocal of beta_2 for optimization simplicity lm.out <-
> > > lm(log(Y)~(X[,2]), weights=as.vector(W)) par <-
> > > c(coefficients(lm.out)[1], 1/coefficients(lm.out)[2]) b0 <- par
> > >
> > > # extract a set of coefficients for initializing the
> > > optimization betahat_ <- optim(par, f, method="CG", X=X, Y=Y,
> > > W=W)$par betahat <- betahat_ betahat[2] <- 1/betahat[2]
> > >
> > > # plot
> > > plot(X[,2], Y, main="Price of IBM vs Time",
xlab="day",
ylab="adjusted
> > price")
> > > lines(X[,2], exp(X %*% betahat), col="blue") b0[2] <- 1/b0[2]
> > > lines(X[,2], exp(X %*% b0), col="red") legend(x=0, y=120,
> > > legend=c("Naive Transformed Least Squares",
"Absolute
Least
Error Predictor"), fill=c("red",
"blue"), bty="n")
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l