hello all,
here are the readings for next week.
Paul W. Holland. 1986. Statistics and Causal Inference. Journal of the
American Statistical Association 81: 945-960.
Paul R. Rosenbaum and Donald B. Rubin. 1985. Constructing a Control
Group Using Multivariate Matched Sampling Methods That Incorporate the
Propensity Score. The American Statistician 39: 33-38.
cheers,
Holger
--
Holger Lutz Kern
Graduate Student
Department of Government
Cornell University
Institute for Quantitative Social Science
Harvard University
1737 Cambridge Street N350
Cambridge, MA 02138
www.people.cornell.edu/pages/hlk23
Hi all,
Below is our working abstract and title.
Best,
Hai and Hea Min
Title:
Explaining the Growth of Bilateral Investment Treaties: Competition and the
Negative Effects of Learning, Past Agreements, and Trade Relations
Abstract:
This paper aims to evaluate Elkins, Guzman, and Simmons (2006)'s
[henceforth, EGS 2006] claim that the spread of bilateral investment
treaties (BITs) derives from the international economic competition between
potential host countries. In short, BITs are intergovernmental treaties
that guarantee particular rights for foreign investors in a host country;
the foreign investors, in turn, invest within the host country to-in
principle-fulfill their contracts' requirements. EGS 2006 additionally
argue that learning based on past or current successful investment
experiences plays a positive role with BIT signing. This paper extends EGS
2006's work by using event-history analysis and accounts for diminishing
marginal benefits and country-level variations. The results show that
competition among host countries does increase the likelihood of signing a
BIT. However, contrary to EGS 2006's findings, this paper demonstrates how
learning, the presence of previous BIT agreements, and dyadic trade
relations between the investor and host country negatively impact the
signing of a BIT. We therefore conclusively argue that the evidence
suggests learning cannot be used to explain the global increase of BITs.
as you can see, i'm being picky. this is because, my experience is that
with my research and that in this class, this is the stage when you can
have the biggest effect. most of the hard work goes into getting the data
up, figuring out the place in the literature, understanding, coding and
running the stat'l model and interpreting the results. but it is after
all that where you can have a big effect if you put the work in to find
the angle, the news, the contribution. and THIS is the time when you have
the infrastructure all in place and so it is far EASIER now than at any
time in doing research to look at one more result, add a new control
variable, look at different subsamples of the time series or for
subsamples of respondents, or calculate a different quantity of interest,
etc... the world will care about the result of that, if you can find a new
angle (and support it); after all, we all know that you worked very hard
and did all the stuff you were supposed to, but that's not the issue and
is not what anyone outside of this class will care about.
Its like the foundation of a building which of course is also absolutely
essential, but once it is done, you burry it and only look at the
beautiful structure built on top of it. don't forget the stuff above
street level...
one question to ask yourself is 'what (or what new) am I claiming?' or as
I've said before 'whose mind will you change about what'. that's what the
abstract should address. as examples, see the abstracts and claims in the
papers on my preprints page: http://gking.harvard.edu/preprints.shtml. I
suggest you just read through them. I mostly write methods papers, and
I'm asking most of you to write about the substance, but other than that
you will see some (attempts to make) unambiguous claims. the success of
the abstract depends on the extent to which you can highlight the new
claim, and the success of the paper depends on the extent to which you can
support it.
Gary
Hello everyone,
I'm partnered with Brian Feinstein and Shauna Shames. We propose the
following title and abstract.
Sincerely,
Tyson
---
Title
Get It Together: Revisiting Tolbert and McNeal's "Unraveling The Effects of
the Internet on Political Participation?"
Abstract
The article entitled, "Unraveling the Effects of the Internet on Political
Participation?" appeared in Political Research Quarterly in June 2003 and
has been subsequently cited 26 times. The analysis used NES survey data to
argue that access to the internet and internet news increases the
probability of voting. If true, this finding has significant consequences
for predicting the future of American politics. Here, we revisit the
original article in order to confirm its results. Unfortunately, we
findthat it can no longer be replicated, even by the authors
themselves.
We therefore aim to retest the article in a fully replicable manner. We
begin by accepting all of the article's hypotheses and regenerating its data
according to the original procedures. We then test the data with an
improved methodology involving matching, multiple imputation, and
simulation. Our exact results differ somewhat from those in the original
article. However, our conclusion remains the same; Internet access and
internet news increase the probability of voting.
Hi all,
We welcome your comments and feedback on our working title and abstract
for our paper.
Thanks,
Sheila and Katy
TITLE:
Scrimping, saving, and skipping pills: A matched analysis of how drug
coverage impacts cost-related nonadherence among elderly Americans.
ABSTRACT:
The goal of this study is to evaluate the relationship between drug
coverage and cost-related nonadherence of medicines among elderly
Americans. We used data from the 2000 wave of the Health and Retirement
Survey to further examine results presented by Mojtabai and Olfson,
2003. We used propensity score matching and a multivariate logit model
to study whether a lack of drug coverage is associated with cost-related
nonadherence. Using unmatched cohorts, Mojtabai and Olfson, (2003) finds
that an increasing levels of drug coverage is associated with increasing
levels of cost-related nonadherence. Our findings indicate that matching
attenuates point estimates but produces larger confidence intervals in
this analysis. The attenuation indicates that selection bias may not be
fully controlled in the original analysis. However, interpretation of
the substantive results does not differ greatly between the original and
matched analyses, due to the overlap in confidence intervals. In the
unmatched cohort the sample average treatment effect estimate is an odds
ratio of 2.91 (2.44, 3.47) and in the matched analysis it is 2.17 (1.72,
2.75). Respondents without drug coverage in the original cohort are
about three times as likely to report cost-related nonadherence, whereas
in matched analysis, these individuals were about twice as likely to
experience cost-related nonadherence as their counterparts with drug
coverage. Similar patterns hold for specific health conditions. This
study provides evidence for the utility of propensity score matching in
observational studies where selection into treatment or exposure is not
random.
Hey Everyone,
There has been some confusion about the abstracts. We would like your
abstracts and titles to be sent out sooner, rather than later.
*At the latest* we'd like for everyone to have circulated their abstracts
by Monday of next week. It would be ideal if these abstracts could be
circulated over the weekend,
Cheers,
Justin
---------- Forwarded message ----------
Date: Wed, 25 Apr 2007 17:31:28 -0400 (EDT)
From: Justin Ryan Grimmer <jgrimmer at fas.harvard.edu>
Reply-To: <gov2001-l at lists.fas.harvard.edu>
To: <gov2001-l at lists.fas.harvard.edu>
Subject: [gov2001-l] 2 Announcements
Hey Everyone,
I hope your projects are going well. With your papers due soon, it is
time to start thinking about your abstract and paper titles. As Gary
notes in "publication, publication" the abstract and title are two of the
most critical components of your paper. Therefore, in order to improve
your abstracts and title, we require that \emph{everyone} send their
abstract and title over the list, so that it can be critiqued by your
fellow classmates, Gary, Holger, and myself.
For more guidance on what you should include in the abstract and title
check out section 5, part (c) of "publication, publication". Also, now
would be a good time to re-read "publication, publication" as we begin the
final push in the class.
Also, for next week's final class please read the following article:
King, Gary. Christopher Murray, Joshua Salomon, and Ajay Tandon. 2004.
"Enhancing the Validity and Cross-Cultural Comparability of Measurement in
Survey Research" (98) 191-207.
We look forward to reading your abstracts and titles!
Cheers,
Justin
_______________________________________________
gov2001-l mailing list
gov2001-l at lists.fas.harvard.eduhttp://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
fyi in case anyone's interested.
Gary
---------- Forwarded message ----------
Date: Fri, 27 Apr 2007 00:01:10 -0400
From: Diana Mutz <mutz at sas.upenn.edu>
To: King at harvard.edu
Subject: TESS SPECIAL CALL DEADLINE JULY 1st, 2007
SPECIAL CALL DEADLINE: July 1st, 2007
ATTENTION ALL SOCIAL SCIENTISTS AND PUBLIC HEALTH RESEARCHERS:
NEW OPPORTUNITIES FOR ORIGINAL DATA COLLECTION VIA TESS!
The National Science Foundation, in cooperation with the Department of
Homeland Security, is funding opportunities for original data collection
for research that has potential relevance to the concerns of DHS. Three
broad substantive areas have been identified as relevant and likely
avenues for research:
1. Risk communication and its effects on disaster preparedness
2. Government and individual attributions of responsibility and
perceived responsiveness; and
3. Inter-group threat and cooperation
Other areas of research with relevance to terrorism, disaster
preparedness, or related public health and medical issues will also be
considered.
Data collection for these projects will be paid for and conducted with
the assistance of Time-sharing Experiments for the Social Sciences
(TESS). As with other TESS applications, only a short, 5 page proposal is
required.
TESS offers researchers opportunities to test their experimental ideas on
large, diverse, randomly selected subject populations, or on specific
subsamples of the population (based on geographic location, demographics,
or other criteria). For more information on this special call or on TESS
opportunities more generally, and for examples of TESS studies completed
in the past, please visit our home page at www.ExperimentCentral.org.
PLEASE NOTE: For those with other areas of interest, TESS continues to
fund data collection for survey-experimental work in all areas of the
social sciences, and for graduate student and faculty investigators in
all disciplines. These proposals are accepted on a rolling basis, and we
will continue accepting new proposals as our funding allows.
Hey Everyone,
I hope your projects are going well. With your papers due soon, it is
time to start thinking about your abstract and paper titles. As Gary
notes in "publication, publication" the abstract and title are two of the
most critical components of your paper. Therefore, in order to improve
your abstracts and title, we require that \emph{everyone} send their
abstract and title over the list, so that it can be critiqued by your
fellow classmates, Gary, Holger, and myself.
For more guidance on what you should include in the abstract and title
check out section 5, part (c) of "publication, publication". Also, now
would be a good time to re-read "publication, publication" as we begin the
final push in the class.
Also, for next week's final class please read the following article:
King, Gary. Christopher Murray, Joshua Salomon, and Ajay Tandon. 2004.
"Enhancing the Validity and Cross-Cultural Comparability of Measurement in
Survey Research" (98) 191-207.
We look forward to reading your abstracts and titles!
Cheers,
Justin
Hi All.
I'm trying to adjust for survey design factors in my reanalysis of replication
data but am unsure about the cluster specification in the 'survey' library. In
particular, I am not sure where/how to incorporate the cluster variable (each
stratum in my sample contains 1-2 clusters).
For anyone familiar with this library - if my cluster variable is 'RAEHSAMP', is
it sufficient to provide the id argument with this variable? For example:
hrs.sbp <- svydesign(id=~RAEHSAMP, strata=hrs.bpmeds$RAESTRAT,
weights=hrs.bpmeds$R5WTRESP, data=hrs.bpmeds)
The example in the help window uses a specification of 'dnum+snum' which makes
little sense to me.
Any clarification would be great. Thanks!
Sheila
Hey Everyone,
the conference below is a great opportunity to get feedback on your work--
hope the papers are going well
Justin
---------- Forwarded message ----------
Date: Mon, 23 Apr 2007 21:23:45 -0400
From: Olivia Lau <olau at fas.harvard.edu>
To: Justin Ryan Grimmer <jgrimmer at fas.harvard.edu>
Subject: Fwd: useR! 2007 online registration is now available
The UseR conference is being held in at Iowa State from August 8-10th
this year. Day-long training courses will be offered on August 8th,
and panels on the 9th and 10th. We have a bunch of interesting
invited speakers lined up, and this is a great chance to get feedback
on your work and interact with individuals from different disciplines.
Both early registration and the deadline for paper submission have
been extended to May 15th. Papers should be about 8 pages, including
graphs and tables -- anything that uses R in an applied manner that
can be generalized beyond the immediate substantive application would
be appropriate (e.g., an edited version most Gov 2001 papers). The
deadline for poster abstracts remains June 30th. While papers will be
refereed, posters will not; if a paper is not accepted for oral
presentation, it will be assigned to a poster session, and all poster
proposals will be assigned to appropriate sessions.
Also, NSF funding is available for graduate students and junior
faculty who are unable to obtain conference funding from their home
institutions. Please email a CV along with your paper or poster
submission to submissions at user2007.org.
Thanks, and hope to see you there!
Olivia Lau
---------- Forwarded message ----------
From: Dianne Cook <dicook at iastate.edu>
Date: Apr 20, 2007 12:33 PM
Subject: useR! 2007 online registration is now available
To: r-announce at stat.math.ethz.ch
Cc: user2007 at iastate.edu
R Users and Developers,
Online registration for useR! 2007 is now available on the conference
web site. The deadline for the early registration discount has been
extended to May 15.
We haven't had many contributed papers as yet, or requests for travel
support from young researchers and graduate students. Please consider
contributing a paper, and encouraging your students and young faculty
to submit a paper and request travel support. The deadline for these
has also been extended to May 15. Contributed papers will be
refereed.
Please also note that we have negotiated airline discounts with both
Northwest and United airlines. They are excellent discounts, there is
some reduction in even the cheapest fares. So do keep this in mind
when you plane your travel. Details are on the web page.
The deadline to submit an abstract for a poster is June 30. These
abstracts will not be refereed. All poster submissions will be
accepted.
Information about the meeting can be found at
http://www.user2007.org/
If you have suggestions for the meeting please email the program
committee at user2007 at iastate.edu.
Douglas Bates, Dianne Cook, Dave Henderson, Heike Hofmann, Olivia Lau,
Michael Lawrence, Luke Tierney, Hadley Wickham
---------------------------
Di Cook
dicook at iastate.edu