Hi all,
Part (b) of the problem set asks to "use Zelig to check
our results." Does "results" refer only to the vector of
parameter estimates or should we also check standard
errors? If the latter is true, then where does Zelig
store the standard errors? (Sorry for not being able to
find the answer in the docs.)
Thank you!
-Alexei
Hi all,
Does anybody knows the mechanics and R command for applying a
Likelihood-ratio test? I found this at wikipedia
http://en.wikipedia.org/wiki/Likelihood-ratio_test, but I do not have a clue
on what command I should use at R. Any ideas?
Best,
Viridiana R?os
617-997-2471
Hi all,
I'm having trouble figuring out how a log-like function maximizing
over multiple parameters would work. It seems for the "par" we would
have to feed the function vectors of guesses for each of the three
betas in the X matrix. How do we choose what to feed into this thing?
It seems difficult to come up with guesses a priori (and unlike a
probability they don't have an obvious range like [0,1])
Thanks,
--
Jon Bischof
Graduate Student
Department of Government
Harvard University
Hi all,
I've uploaded a brief but more thorough explanation of expected and
predicted values to the section folder on the course website. I composed
the comments in an R file thinking you may find it helpful to copy and paste
it into your script or into the section script.
There are a couple of ways to think about predicted values, and part of the
goal of the problem set is to get you thinking about that. If in doubt,
explain what you did and why you think it makes sense.
Happy St. Patrick's Day Eve,
Jenn
I'm having problems implementing the log likelihood function. I used
the following code:
loglogit <- function(par, X=X, y=y){
beta <- par[1:ncol(X)]
a <- (X*beta)
b <- (1 - (2*y))
c <- 1 + exp(b%*%a)
d <- log(c)
out <- sum(d)
}
And when I give it to optim I get very unlikely coefficients and a
Hessian of all zeros. I can't seem to figure out what I am
overlooking. Does anybody see any obvious problems with this
function? If so, I would appreciate a nudge in the right direction.
Thanks,
Keith
Hi all,
I gather there is some confusion brewing about simulating expected and
predicted values. Tonight I will add more comments to the section code and
re-post it to try to assist.
Just remember- thinking about these sorts of things is character-building.
Jenn
Hi Viridiana,
The terms 'expected' and 'predicted' as presented on the lecture slides
refer to the value of the outcome variable y, so we are interested in the
difference between expected values and predicted values.
In this case, an expected value is a probability, so we can use the terms
expected probability and expected value interchangeably. Notice that if we
average over a lot of outcome variables, we're averaging a lot of 1's and
0's, which means we're dividing the total number of 1's by the total number
of 1's and 0's, which means we get a probability of o-ring failure. (hence,
the last step in simulating an expected value returns a probability).
You rightly point out that the dependent variable in this case is
dichotomous- either o-rings fail or they do not; y can only be a zero or a
one. To estimate one expected value, you draw one set of betas (a B0 and a
B1), fix your x's, and calculate a probability using the function you
mentioned (1/(1+e^(-Bxi)) ). That function will return n probabilities,
where n is the number of observations. Plug those n probabilities in to the
stochastic component to get n y's, average those n y's, and you have one
expected value/expected probability. (note that these last two steps are
superfluous in the case of a logit model).
Hope this helps,
Jenn
On Sat, Mar 15, 2008 at 3:33 PM, Viridiana Rios <vrios at fas.harvard.edu>
wrote:
> Hi all,
>
> Can anybody please, explain to me the difference between expected values
> and expected probability in the context of a dichotomic dependent variable?
>
> Here is my confusion: We have a probit model (y=B0+B1x1). We draw 1000
> Betas from a normal distribution. We calculate y=B0+B1x1 using all the Betas
> of the draw. So, now we have 1000 different values of y (one for each draw
> of Betas). Then, we transform y to a probability using the function
> 1/(1+e^(-Bxi)). Now we have 1000 different probabilites. Are these
> probabilities the expected probability or the expected value? If these are
> the expected probabilities, howe can I get the expected values? Are the
> expected values the 1000 values of y?
>
> Best,
>
> Viridiana R?os
> 617-997-2471
>
Hi all,
Can anybody please, explain to me the difference between expected values and
expected probability in the context of a dichotomic dependent variable?
Here is my confusion: We have a probit model (y=B0+B1x1). We draw 1000 Betas
from a normal distribution. We calculate y=B0+B1x1 using all the Betas of
the draw. So, now we have 1000 different values of y (one for each draw of
Betas). Then, we transform y to a probability using the function
1/(1+e^(-Bxi)). Now we have 1000 different probabilites. Are these
probabilities the expected probability or the expected value? If these are
the expected probabilities, howe can I get the expected values? Are the
expected values the 1000 values of y?
Best,
Viridiana R?os
617-997-2471
---------- Forwarded message ----------
Date: Fri, 14 Mar 2008 11:58:09 -0400
From: "Franzese, Robert" <franzese at UMICH.EDU>
Reply-To: Political Methodology Society <POLMETH at ARTSCI.WUSTL.EDU>
To: POLMETH at ARTSCI.WUSTL.EDU
Subject: [POLMETH] REMINDER: PolMeth XXV: Applications Due in ONE Week!!!!
Hi Folks,
Just another pestering reminder about PolMeth XXV:
The deadline for applications to the conference is now in 1 week!!!
20 March 2008
To apply to attend the meetings, complete the application form at the
Society's webpage by 20 March 2008:
http://polmeth.wustl.edu/conferences/methods2008/register/
Submit a paper proposal, a poster proposal, a proposal to attend, and/or
a proposal indicating your willingness to serve as discussant.
Further Reminders:
We have expanded the planned size of the conference, so there
will be plenty of space for a much higher percentage of applicants to
attend and, if the supply of proposals suffices, of poster/paper
proposals to be accepted.
Faculty may propose to present in poster or paper sessions and,
if they propose papers, may be assigned nonetheless to a
poster-presentation at the conference-organizers sole discretion.
(Almost assuredly, if you propose a poster, you will get that format if
accepted.)
Graduate students are urged to propose a poster for the
graduate-student poster-session, and those who do will receive higher
priority for the limited funding we have, but students may also propose
simply to attend.
So: APPLY!
DEADLINE: 20 March 2008
Apologies for the clutter &/or cross-postings.
Looking forward to seeing you there!
Rob
P.S. Did I mention? DEADLINE: 20 March 2008
The original call follows:
-----Original Message-----
From: franzese at umich.edu [mailto:franzese at umich.edu]
Sent: Tuesday, February 19, 2008 8:39 AM
To: Franzese, Robert
Subject: PolMeth XXV: Call for Applications to Participate
CALL FOR PROPOSALS:
PolMeth XXV: The 25th Annual, Silver Edition
Summer Meeting of the Society for Political Methodology
The University of Michigan, Ann Arbor
9 - 13 July 2008 (Core Program: 10-12 July 2008)
The 25th Annual, Silver Edition, Summer Methodology Conference will be
held 10-12 July 2008, on the campus of the University of Michigan, Ann
Arbor. Sponsors of these meetings are the Institute for Social Research
of
the University of Michigan, the National Science Foundation, and the
Society for Political Methodology.
To apply to attend the meetings, complete the application form at the
Society's webpage by 20 March 2008:
http://polmeth.wustl.edu/conferences/methods2008/register/
DEADLINE FOR APPLICATIONS: 20 March 2008.
All applicants will propose to present a poster or a paper, and/or to
act
as a discussant, or simply to attend. Although graduate students may
also
apply simply to attend, they are strongly encouraged to propose a poster
for the graduate-student poster-session, which is a venerated tradition
and outstanding professional opportunity offered by the conference.
Graduate students (and other eligible applicants) proposing and having
accepted a poster (or a paper) will receive priority consideration for
funding if they apply for it (see below).
To accommodate the continued growth of the conference in response to
great
and strongly rising demand--we expect a further approximately 50%
increase
in size from last year's expanded total of about 160 attendees--we have
made a number of small adjustments in the format of the conference,
requiring corresponding alterations of the application and registration
processes.
SUMMARY OF THE REVISED PROGRAM-FORMAT (much further information
available
at the conference web-site:
http://polmeth.wustl.edu/conferences/methods2008/):
The conference program will consist of:
One day (Thursday, 10 July) of plenary sessions (i.e., sessions of the
whole) of paper presentations, with discussant & open discussion.
One day (Friday, 11 July) of split-sessions, split in two or possibly
three, of paper presentations with discussant & open discussion or of
poster session(s), but not both types simultaneously.
One day (Saturday, 12 July) of split-sessions, split in two, of paper
presentations with discussant & open discussion or of poster session(s),
but not both types simultaneously.
Also planned for Thursday: the traditional keynote lecture by a
prominent, non-political-science, methodologist from U of M, and our
business meeting, followed by a opening-evening reception.
Also planned for Friday: the traditional graduate-student
poster-session, accompanied by an evening reception.
Also planned for Saturday: a closing dinner-reception.
A new tradition: NSF has funded also a dinner, reception, and/or other
event for Women in Political Methodology, tentatively scheduled for
Wednesday evening. (This will be the third such event.)
An even newer tradition: An add-on mini-conference/workshop or two may
be held Wednesday PM and/or Sunday AM, on topics and of a format TBA.
Announcements as those develop will go to this email list and the
conference web site.
The core conference begins Thursday morning and ends Saturday evening,
so
attendees should plan to arrive Wednesday the 9th and depart Sunday the
13th.
To accommodate the expanded size, we have increased the numbers of split
sessions, and, also partly in response to that increased size, we intend
to expand usage of the poster-session format. Poster presentations and
split-session paper-presentations will be scheduled and treated as full
equals in every regard with plenary paper-presentation sessions.
Applicants may propose a poster or a paper, but the program committee
will
assign accepted paper-proposals to plenary-sessions, to split-session
paper-presentations, or to poster-session presentations entirely on
their
own discretionary judgment regarding optimal conference design, balance,
and flow.
NOTABLE REGARDING THE REVISED APPLICATION & REGISTRATION PROCESSES:
Following acceptance and upon registration, all conference attendees
will
pay or have paid for them the $200 conference registration fee. (This is
the same amount as the faculty-registration fee last year, but now all
attendees will pay it. This is also approximately equal to the variable
cost per person of the conference.) The fee is collected in the
registration process (via PayPal), and registration cannot be validly
completed without payment.
Newly expanded this year, the NSF will fund the attendance of 55 (total)
graduate students, women, minorities, and assistant professors. (Great &
sincere gratitude is hereby extended from us all to PI's Jan
Box-Steffensmeier & Phil Schrodt.) The acceptance committee will review
applications for funding (which, procedurally, amount to the applicant
indicating her/his wish to be considered for funding and then checking
boxes for the categories by which s/he is eligible), and winners of
funding will be notified along with their acceptance to the conference.
Also new this year, all graduate-student applicants will be required to
give the email address not only of a faculty recommender, which has long
been required, but also to give the email address of a "guarantor of
funding." The guarantor may be the same faculty member as the
recommender,
a different faculty member, someone authorized to commit an
institutional
source of funding, or even the graduate student her/himself. The
guarantor
commits to paying or having paid the registration fee for the graduate
student in the case that student does not win funding. The option of the
student listing her/himself as guarantor allows students wishing (or
able
and willing and needing) to pay their own registration to do so. The
options, purposefully listed first, of naming a faculty or institutional
guarantor are intended to emphasize heavily the strong and good norm
that
faculty/departmental/institutional sponsors should pay for their
students
to attend the conference.
ANNOUNCEMENT OF ACCEPTANCES & OF FUNDING SHOULD ARRIVE AROUND MID-APRIL.
Following that, LODGING _MUST_ BE BOOKED/CONFIRMED BY END OF MAY.
Technical questions regarding the application & registration web-site
should be directed to Stephen Haptonstahl (srhapton at wustl.edu).
Substantive questions or comments regarding the application process, the
conference, or the conference web-site should be directed to Rob
Franzese
(franzese at umich.edu) or the graduate-student assistant for the
conference
& conference web-site, Bryce Corrigan (becorrig at umich.edu).
We look forward to seeing you in Ann Arbor this summer!
************************************************************************
*
Robert (Rob) J. Franzese, Jr. US Mail: (Room 4246
ISR)
Associate Professor, P.O. Box
1248
Department of Political Science, Ann Arbor, MI
48106-1248
and Research Associate Professor, (Currier: 426 Thompson
St.)
Center for Political Studies, e-mail:
franzese at umich.edu
Institute for Social Research, office:
1-734-936-1850
The University of Michigan fax:
1-734-764-3341
http://www-personal.umich.edu/~franzese
************************************************************************
*
**********************************************************
Political Methodology E-Mail List
Editors: Melanie Goodrich, <melaniegoodrich at nyu.edu>
Delia Bailey, <dbailey at wustl.edu>
**********************************************************
Send messages to polmeth at artsci.wustl.edu
To join the list, cancel your subscription, or modify
your subscription settings visit:
http://polmeth.wustl.edu/polmeth.php
**********************************************************
**********************************************************
Political Methodology E-Mail List
Editors: Melanie Goodrich, <melaniegoodrich at nyu.edu>
Delia Bailey, <dbailey at wustl.edu>
**********************************************************
Send messages to polmeth at artsci.wustl.edu
To join the list, cancel your subscription, or modify
your subscription settings visit:
http://polmeth.wustl.edu/polmeth.php
**********************************************************