
1 
Paper

Bayesian exploratory data analysis
Gelman, Andrew

Uploaded 
02112003

Keywords 
bootstrap Fisher's exact test graphics mixture model model checking multiple imputation prior predictive check posterior predictive check pvalue uvalue

Abstract 
Exploratory data analysis (EDA) and Bayesian inference (or, more
generally, complex statistical modeling)which are generally
considered as unrelated statistical paradigmscan be particularly
effective in combination. In this paper, we present a Bayesian
framework for EDA based on posterior predictive checks. We explain
how posterior predictive simulations can be used to create reference
distributions for EDA graphs, and how this approach resolves some
theoretical problems in Bayesian data analysis. We show how the
generalization of Bayesian inference to include replicated data $y^{
m
rep}$ and replicated parameters $ heta^{
m rep}$ follows a long
tradition of generalizations in Bayesian theory.
On the theoretical level, we present a predictive Bayesian formulation of
goodnessoffit testing, distinguishing between $p$values (posterior
probabilities that specified antisymmetric discrepancy measures will
exceed 0) and $u$values (data summaries with uniform sampling
distributions). We explain that $p$values, unlike $u$values, are
Bayesian probability statements in that they condition on observed data.
Having reviewed the general theoretical framework, we discuss the
implications for statistical graphics and exploratory data analysis, with
the goal being to unify exploratory data analysis with more formal
statistical methods based on probability models. We interpret various
graphical displays as posterior predictive checks and discuss how
Bayesian inference can be used to determine reference distributions.
The goal of this work is not to downgrade descriptive statistics, or to
suggest they be replaced by Bayesian modeling, but rather to suggest how
exploratory data analysis fits into the probabilitymodeling paradigm.
We conclude with a discussion of the implications for practical Bayesian
inference. In particular, we anticipate that Bayesian software can be
generalized to draw simulations of replicated data and parameters from
their posterior predictive distribution, and these can in turn be used to
calibrate EDA graphs. 

2 
Paper

Time Series Models for Compositional Data
Brandt, Patrick T.
Monroe, Burt L.
Williams, John T.

Uploaded 
07091999

Keywords 
compositional data VAR time series analysis bootstrap Monte Carlo simulation macropartisanship

Abstract 
Who gets what? When? How? Data that tell us who got what are compositional data
 they are proportions that sum to one. Political science is, unsurprisingly,
replete with examples: vote shares, seat shares, budget shares, survey marginals,
and so on. Data that also tell us when and how are compositional time series data.
Standard time series models are often used, to detrimental consequence, to model
compositional time series. We examine methods for modeling compositional data
generating processes using vector autoregression (VAR). We then use such a method
to reanalyze aggregate partisanship in the United States. 

4 
Paper

Does Size Matter? Exploring the Small Sample Properties of Maximum Likelihood Estimation
Hart, Jr., Robert A.
Clark, David H.

Uploaded 
04201999

Keywords 
small samples ML Type II errors bootstrap

Abstract 
The last two decades have witnessed an explosion in the use of
computationally intensive methodologies in the social sciences as
computer technology has advanced. Among these empirical methods are
Maximum Likelihood (ML) procedures. ML estimators possess a variety
of desirable qualities, perhaps most prominent of which is the
asymptotic efficiency of the standard errors. However, the behavior
of the estimators in general, of the estimates of the standard errors
in particular, and thus of inferential hypothesis tests are uncertain
in small sample analyses. In political science research, small
samples are routinely the subject of empirical investigation using ML
methods, yet little is known regarding what effect sample size has on
a researcher's ability to draw inferences
This paper explores the behavior of ML estimates in probit models
across differing sample sizes and with varying numbers of independent
variables in Monte Carlo simulations. Our experimental results allow
us to conclude that: a) the risk of making Type I errors does not
change appreciably as sample size descends; b) the risk of making Type
II errors increases dramatically in smaller samples and as the number
of regressors increases. 

6 
Paper

Bootstrap Methods for Nonnested Hypothesis Tests
Mebane, Walter R.
Sekhon, Jasjeet

Uploaded 
07201996

Keywords 
Cox Test Bootstrap LISREL Endogenous Switching Regression TobitStyle Censoring

Abstract 
Cox (1961; 1962) proposed a fairly general method that can be used to
construct powerful tests of alternative hypotheses from separate statistical
families. We prove that nonparametric bootstrap methods can produce
consistent and secondorder correct approximations to the distribution of the
Cox statistic for nonnested LISRELstyle covariance structure models. We use
the method to investigate a question about the specification of a LISREL model
used by Kinder, Adams and Gronke (1989). In a second applicationa pair of
nonnested endogenous switching regression models with tobitstyle censoring,
applied to real datawe illustrate how bootstrap calibration can be used to
correct the size of the test when the test distribution is being estimated by
Monte Carlo simulation due to concern about nonregularity. 

11 
Paper

NonCompulsory Voting in Australia?: what surveys can (and can't) tell us
Jackman, Simon

Uploaded 
08251997

Keywords 
turnout Australian politics compulsory voting political participation counterfactuals surveys nonresponse measurement error socialdesirability heuristic questionorder effects simulation parametric bootstrap

Abstract 
Compulsory voting has come under close scrutiny in recent Australian
political debate, and influential voices within the (conservative)
Coalition government have called for its repeal. Conventional
wisdom holds that a repeal of compulsory voting would result in a
sizeable electoral boost for the Coalition; the proportion of
Coalition voters who would not vote is thought to be smaller than
the corresponding proportion of Labor voters. But estimates of
Coalition gains under a return to voluntary turnout are quite
roughandready, relying on methods hampered by critical
shortcomings. In this paper I focus on assessing the counterfactual
of noncompulsory turnout via surveys: while turnout is compulsory
in Australia, responding to surveys isn't, and the problems raised
by high rates of nonresponse are especially pernicious in
attempting to assess the counterfactual of voluntary turnout.
Among survey respondents, socialdesirability and questionorder
effects also encourage overreports of the likelihood of voluntarily
turning out. Taking nonresponse and measurement error into
consideration, I conclude that surveybased estimates (a)
significantly emph{underestimate} the extent to which turnout
would emph{decline} under a voluntary turnout regime; but (b)
emph{overestimate} the extent to which a fall in turnout would
work to the advantage of the Coalition parties. Nonetheless, the
larger of the Coalition parties  the Liberal Party 
unambiguously increases its vote share under a wide range of
assumptions about who does and doesn't voluntarily turnout. 

12 
Paper

The Changing Economic Preferences of the American Public: 19761991
Sekhon, Jasjeet

Uploaded 
03281997

Keywords 
Macroeconomic Preferences Monetary Policy Hermite Polynomials Splines Bootstrap

Abstract 
I show that the public indeed does have coherent preferences over
macroeconomic tradeoffs, and these preferences have changed in ways
consistent with not only economic theory but also with the changes
which occurred in the American political system during the 1980s. In
particular, most people learned something new about the state of the
world in the late 1970s, and began to reject classical Keynesian
explanations about economic reality. Individuals were becoming more
sympathetic to the economic platform of the Republican partyi.e.,
they began to favour price stability. Moreover, the results support
the notion that poor Americans do not hold government policy
responsible for their personal economic plight (Hochschild 1981, Lane
1962). 

13 
Paper

Data Mining for Theorists
Kenkel, Brenton
Signorino, Curtis

Uploaded 
07262011

Keywords 
empirical implications of theoretical models basis regression adaptive lasso bootstrap functional form misspecification

Abstract 
Among those interested in statistically testing formal models,two approaches dominate. The structural estimation approach derives a structural probability model based on the formal model and then estimates parameters associated with that model. The reducedform approach generally applies offtheshelf techniquessuch as OLS, logit, or probitto test whether the independent variables are related to a decision variable according to the comparative statics predictions. We provide a new statistical method for the comparative statics approach. The decision variable of interest is modeled as a polynomial function of the available covariates, which allows for the nonmonotonic and interactive relationships commonly found in strategic choice data. We use the adaptive lasso to reduce the number of parameters and prevent overfitting, and we obtain measures of uncertainty via the nonparametric bootstrap. The method is "data mining" because the aim is to discover complex relationships in data without imposing a particular structure,but "for theorists" in that it was developed specifically to deal with the peculiar features of data on strategic choice. Using a Monte Carlo simulation, we show that the method handily outperforms other nonstructural techniques in estimating a nonmonotonic relationship from strategic choice data. 

14 
Paper

Unresponsive, Unpersuaded: The Unintended Consequences of Voter Persuasion Efforts
Bailey, Michael
Hopkins, Daniel
Rogers, Todd

Uploaded 
08092013

Keywords 
causal inference field experiments persuasion attrition multiple imputation Approximate Bayesian Bootstrap

Abstract 
Can randomized experiments at the individual level help assess the persuasive effects of campaign tactics? In the contemporary U.S., vote choice is not observable, so one promising research design to assess persuasion involves randomizing appeals and then using a survey to measure vote intentions. Here, we analyze one such field experiment conducted during the 2008 presidential election in which 56,000 registered voters were assigned to persuasion in person, by phone, and/or by mail. Persuasive appeals by canvassers had two unintended consequences. First, they reduced responsiveness to the followup survey, lowering the response rate sharply among infrequent voters. Second, various statistical methods to address the resulting biases converge on a counterintuitive conclusion: the persuasive canvassing reduced candidate support. Our results allow us to rule out even small effects in the intended direction, and illustrate the backlash that persuasion can engender. 

15 
Poster

The Impact of Sampling Procedures on Statistical Inference with Clustered Data
Jin, Shuai

Uploaded 
07202015

Keywords 
Sampling Clustering Monte Carlo Jackknife Bootstrap

Abstract 
This study explores the performances of multiple methods handling clustering under different sampling procedures. Many disciplines have proposed various methods to correct the downward bias in the OLS variance estimates with clustering. However, these methods do not take into account the effects of sampling procedures. The sampling procedures affect how clustering in the population enters into the samples; therefore, it affects the performance of the methods analyzing clustered data. This study compares eight methods of estimating variances of linear regression coefficients with clustered data under three different sampling procedures. Monte Carlo simulation results show that sampling procedures affect the variance estimates of both the grouplevel and the individuallevel independent variables. Simple random sampling produces stable and small standard errors. Jackknife cluster standard errors generally perform well. This study analyzes a national Chinese survey dataset in the application section. The results from the real data confirm the conclusions from the Monte Carlo simulations. 

