Estimating the causal effect of dynamic treatment regimes in observational
studies. Miguel Hernan, MD, Instructor in Epidemiology, Harvard School of Public Health. Joint work
with James Robins and Stephen Cole
Wednesday November 13 at noon
Center for Basic Research in the Social Sciences
34 Kirkland Street, Room 22
Lunch will be served.
Abstract:
Large randomized experiments with full adherence and no losses to
follow-up can be used to make causal inferences without further
assumptions. These ideal experiments may be used to evaluate the effect of
treatment regimes that a) fully describe each subject's treatment
experience during the study even before the study starts (e.g., always
treat, never treat, treat every other month), or b) do not fully describe
each subject's treatment experience during the study before the study
starts because the treatment actually received depends on the subject's
time-varying characteristics (e.g., treat until side effects appear, treat
only when some blood parameter reaches the value 200). We call the latter
dynamic treatment regimes. The analysis of ideal experiments is
straightforward: simply compare the mean outcome (or survival) in subjects
under each (dynamic or non-dynamic) treatment regime.
Observational studies can be used to make causal inferences only under the
assumption that they are able to provide the same results as a (possibly
hypothetical) randomized experiment. This assumption, also known as "the
assumption of no unmeasured confounders" or "of sequential randomization",
is not testable. But even if this assumption holds, and in the absence of
model misspecification, the standard analysis of observational studies
with time-varying treatments may yield estimates that cannot be endowed
with a causal interpretation. This limitation of standard methods applies
to the estimation of the effects of both dynamic and non-dynamic treatment
regimes. In contrast, methods based on marginal structural models and
structural nested models would yield valid causal estimates.
We have previously applied a marginal structural Cox model to estimate the
effects of non-dynamic treatments in HIV/AIDS epidemiology. The
subject-matter question we addressed was "should HIV-infected individuals
be treated with highly active antiretroviral therapy?" In this talk, I
will review our previous results and describe the application of a
structural nested accelerated failure time model for estimating the effect
of dynamic treatments among HIV-infected individuals. The question is
"should treatment with highly active antiretroviral therapy start when CD4
count drops to 200 or to 350?" Also, I will discuss computational issues
and different methods to handle censoring.
When the paper becomes available, a link will be sent to those requesting
it (email corr(a)fas.harvard.edu) and to those on the workshop email list.
Those unfamiliar with methods for causal inference from observational data
may find it useful to read about estimating non-dynamic treatments as
described in
http://www.hsph.harvard.edu/causal/publications/joint-causal.pdf
The Research Workshop in Applied Statistics is a forum for graduate
students, faculty, and visiting scholars to present and discuss
statistical innovations and applications in the social sciences. For more
information or to subscribe to the permanent workshop list, email
corr(a)fas.harvard.edu.
Paper now available at:
http://www.courses.fas.harvard.edu/~gov3009/handouts/comovementinternet.pdf
Shrewd, Crude, or Simply Deluded? Market Classification and the Internet
Stock Phenomenon. Ezra Zuckerman (MIT, Sloan School of Management)
Wednesday November 6 at noon
Center for Basic Research in the Social Sciences
34 Kirkland Street, Room 22
Lunch will be served.
Abstract:
We analyze comovement among Internet and other categories of stocks during
the late 1990s and 2000 in an effort to assess the sophistication of
stock-market valuation. Prominent accounts of the Internet stock
phenomenon suggest that the prices of these stocks were determined by
simplistic thinking. In particular, investors were not discriminating as
they crudely grouped all Internet stocks into an undifferentiated and
highly attractive investment category. We find that, in fact, comovement
among Internet stocks was high throughout much of this period but did not
reach the very high levels assumed by prevailing accounts. In addition, we
describe two additional patterns that are problematic for such
interpretations. First, comovement is less characteristic of price
increases than of price drops. Second, Internet stocks exhibited moderate
to high period-to-period consistency in the manner by which category
members were distinguished by investors. Together, our evidence supports
our view that, rather than being anomalous, the Internet stock phenomenon
was symptomatic of the general way by which equity prices are determined:
valuation is driven by prevailing theories of value, which are reasonable
but quite fallible. This view has important implications for how scholars
and managers understand and react to stock market dynamics.
The paper will be posted as soon as it becomes available on the workshop
website, www.courses.fas.harvard.edu/~gov3009/fall02/
The Research Workshop in Applied Statistics is a forum for graduate
students, faculty, and visiting scholars to present and discuss
statistical innovations and applications in the social sciences. For more
information, contact corr(a)fas.harvard.edu.
Shrewd, Crude, or Simply Deluded? Market Classification and the Internet
Stock Phenomenon. Ezra Zuckerman (MIT, Sloan School of Management)
Wednesday November 6 at noon
Center for Basic Research in the Social Sciences
34 Kirkland Street, Room 22
Lunch will be served.
Abstract:
We analyze comovement among Internet and other categories of stocks during
the late 1990s and 2000 in an effort to assess the sophistication of
stock-market valuation. Prominent accounts of the Internet stock
phenomenon suggest that the prices of these stocks were determined by
simplistic thinking. In particular, investors were not discriminating as
they crudely grouped all Internet stocks into an undifferentiated and
highly attractive investment category. We find that, in fact, comovement
among Internet stocks was high throughout much of this period but did not
reach the very high levels assumed by prevailing accounts. In addition, we
describe two additional patterns that are problematic for such
interpretations. First, comovement is less characteristic of price
increases than of price drops. Second, Internet stocks exhibited moderate
to high period-to-period consistency in the manner by which category
members were distinguished by investors. Together, our evidence supports
our view that, rather than being anomalous, the Internet stock phenomenon
was symptomatic of the general way by which equity prices are determined:
valuation is driven by prevailing theories of value, which are reasonable
but quite fallible. This view has important implications for how scholars
and managers understand and react to stock market dynamics.
The paper will be posted as soon as it becomes available on the workshop
website, www.courses.fas.harvard.edu/~gov3009/fall02/
The Research Workshop in Applied Statistics is a forum for graduate
students, faculty, and visiting scholars to present and discuss
statistical innovations and applications in the social sciences. For more
information, contact corr(a)fas.harvard.edu.
Methodologist as Arbitrator: Five Models for Black-White Differences in
the Causal Effect of Expectations on Attainment. Professor Stephen L.
Morgan (Sociology, Cornell University)
Wednesday October 30 at noon
Center for Basic Research in the Social Sciences
34 Kirkland Street, Room 22
Lunch will be served.
Abstract:
When progress in applied research slows because opposing coalitions of
investigators privilege their favored models, methodologists can contribute
by addressing a tractable unresolved question that is relevant to all
competing positions. In this article the literature on educational
attainment is addressed broadly by focusing on alternative positions on
the need to model students' own beliefs and more narrowly by attempting to
answer a classic question that emerged in debates over the power of status
attainment approaches: Why is the relationship between educational
expectations and subsequent educational attainment weaker for blacks than
for whites? Analyzing the High School & Beyond data, five complementary
models of the causal effect of expectations on attainment are offered: a
traditional path-model, an average effects instrumental variable model, a
counterfactual analysis of bounds, a rational expectations forecasting
model, and a panel data model of updated expectations. The general
methodological goal is to demonstrate how multi-model research can be
pursued: After delineating the motivating assumptions of alternative
theoretical positions and the explanations that they sustain, one then
simultaneously estimates a wide range of plausible models, derives
alternative permissible conclusions, and specifies the implied research
program that must be mounted to choose from among the plausible
conclusions.
The paper may be accessed at:
http://www.courses.fas.harvard.edu/~gov3009/handouts/methodologistoct2002.p…
The Research Workshop in Applied Statistics is a forum for graduate
students, faculty, and visiting scholars to present and discuss
statistical innovations and applications in the social sciences. For more
information or to receive workshop announcements every week, contact
corr(a)fas.harvard.edu.
Randomness and Coincidences: Reconciling Intuition and Probability
Theory
Presentation by Tom Griffiths (Stanford University, Psychology
Department).
Coauthored by Joshua Tennenbaum (Stanford University, Psychology
Department).
Wednesday October 23 at noon
Center for Basic Research in the Social Sciences
34 Kirkland Street, Room 22
Lunch will be served.
Abstract:
People are notoriously bad at reasoning about chance. Our intuitions about
randomness and coincidences seem to be inconsistent with the normative
structure of statistics: events that are equally likely to arise by chance
differ in their subjective randomness, and we consistently underestimate
the probability of coincidences. I will argue that we can understand why
these apparent errors arise by focusing on the evidence observations
provide about the processes that generated them rather than their
probability of occurring by chance. As part of this argument, I will
present a simple Bayesian framework that gives predictions about both the
extent to which a set of events will seem random, and the strength of
coincidences. This framework accurately predicts behavior in a variety of
contexts, suggesting that statistics might yet contribute to understanding
our intuitions about chance.
The Research Workshop in Applied Statistics is a forum for graduate
students, faculty, and visiting scholars to present and discuss
statistical innovations and applications in the social sciences. For more
information, contact corr(a)fas.harvard.edu.
The paper may be accessed at:
http://www-psych.stanford.edu/~gruffydd/papers/random.pdf
The presenter's email is:
gruffydd(a)psych.stanford.edu
"Geography, Power, and the Size of Nations: An Agent-Based Model of the
International System"
Presentation by Anders Corr (Harvard University, Government Department)
Wednesday October 16 at noon
Center for Basic Research in the Social Sciences
34 Kirkland Street, Room 22
Lunch will be served.
The Research Workshop in Applied Statistics is a forum for graduate
students, faculty, and visiting scholars to present and discuss
statistical innovations and applications in the social sciences. For more
information, contact corr(a)fas.harvard.edu.
Abstract:
Power discrepancy is an observed characteristic of the international
system. Lars-Erik Cederman's agent-based computer model, GeoSim, simulates
the interaction of states in an international system, but produces
equilibria in which power is distributed homogenously over system members
(Cederman 2002). Alesina and Spolaore (forthcoming) likewise model changes
in the size of nations without accounting for equilibria in which small
and large states coexist. This study enriches Cederman's model by adding
geography, thereby supplying a missing link in both literatures. As
expected, geography yields a long-run distribution of power consistent
with the empirical discrepancy between states, including the simultaneous
existence of small and hegemonic polities.
The paper may be accessed at:
http://www.courses.fas.harvard.edu/~gov3009/handouts/geosimgeog8.pdf
``Estimation of Optimal Treatment or Sequential Decision Strategies from
Non-Experimental Data Using Structural Nested Models"
Presentation by Professor James Robins (Harvard University, School of
Public Health)
Wednesday October 9 at noon
Center for Basic Research in the Social Sciences
34 Kirkland Street, Room 22
Lunch will be served.
The Research Workshop in Applied Statistics is a forum for graduate
students, faculty, and visiting scholars to present and discuss
statistical innovations and applications in the social sciences. For more
information, contact corr(a)fas.harvard.edu.
This week's paper is not available prior to the presentation.
"The importance of statistical methodology for analyzing data from field
experimentation: Evaluating voter mobilization strategies" by Kosuke Imai
(Government, Harvard University)
The first presentation of the Research Workshop in Applied Statistics
will be held on Wednesday 9/25 at noon, CBRSS Rm. 22. Lunch will be
served. The paper may be accessed at:
http://www.courses.fas.harvard.edu/~gov3009/fall02/
Abstract:
Field experimentation is making its way back into the toolkit of political
scientists. Gerber and Green have led this important methodological
development that is likely to improve causal inferences in political
science research. However, they believe that field experiments only
require "rudimentary data analysis." Countering this claim, I use Gerber
and Green's voter mobilization data (2000) to show that statistical
methods are essential to address complications that invariably arise in
field experiments. I demonstrate how incomplete randomization of treatment
assignment led to the authors' puzzling finding that get-out-the-vote
calls discourage voters from going to the polls, reducing turnout by 5
percent. An application of matching, which is more appropriate given the
incomplete randomization, reveals that telephone canvassing increases
turnout by about 5 percent. My analysis also finds that mail canvassing is
a signficant cost-effective alternative, and that appeals related to civic
engagement are more effective than the original analysis indicated.
Contact: corr(a)fas.harvard.edu
***************************************
Research Workshop in Applied Statistics
***************************************
Jasjeet Sekhon (Government)
Garrett FitzMaurice (School of Public Health)
Lee Fleming (Business School)
Gary King (Government)
Donald Rubin (Statistics)
Christopher Winship (Sociology)
The Research Workshop in Applied Statistics is a forum for graduate
students, faculty, and visiting scholars to present and discuss work
in progress and exchange ideas. It is intended as a tour of Harvard's
statistical innovations and applications with weekly stops in
different disciplines such as economics, epidemiology, medicine,
political science, psychology, public policy, public health, sociology and
statistics. The topics of papers presented in previous years included
missing data, survey analysis, Bayesian simulation, sample selection,
and models for election and portfolio choice. Faculty and student
participants in the workshop present their current projects, and guest
speakers also give occasional presentations. The workshop provides an
excellent opportunity for informal interaction between graduate
students and faculty from a variety of disciplines. Course credit is
available for students as Government 3009. Lunch is provided.
If you are interested, please be sure to attend our first meeting on
Wednesday, September 18th at noon, in Room 22, Center for Basic Research
in Social Sciences (CBRSS, 34 Kirkland St., this is the yellow building
across the street from William James Hall). Contact information
and previous presentations may be found at the course web site:
http://www.courses.fas.harvard.edu/~gov3009/
To join the gov3009 mailing list, send e-mail to
gov3009-l-request(a)fas.harvard.edu with the following text message:
subscribe
end
Questions? Please contact the workshop coordinator, Anders Corr, at
corr(a)fas.harvard.edu
*********************************************************************
*********************************************************************
Hi Everyone,
This Wednesday, April 24th, David van Dyk and Kosuke Imai will be
presenting their paper entitled, "Causal inference with general treatment
regimes".
The paper can be downloaded from our course page
(http://www.courses.fas.harvard.edu/~gov3009/spring02)
or from
(http://www.people.fas.harvard.edu/~kimai/files/pscore.ps).
See you all Wednesday!
Shigeo
__________________________________________________________________
Shigeo Hirano Political Economy and Government
shirano(a)fas.harvard.edu Harvard University