logoPicV1 logoTextV1

Search Results


Below results based on the criteria 'bayesian'
Total number of records returned: 77

1
Paper
Estimation and Inference by Bayesian Simulation: an on-line resource for social scientists
Jackman, Simon

Uploaded 08-30-1999
Keywords Markov chain Monte Carlo
Bayesian statistics
how-to
BUGS
ordinal probit
time series
Abstract http://tamarama.stanford.edu/mcmc a Web-based on-line resource for Markov chain Monte Carlo, specifically tailored for social scientists. MCMC is probably the most exciting development in statistics in the last ten years. But to date, most applications of MCMC methods are in bio-statistics, making it difficult for social scientists to fully grasp the power of MCMC methods. In providing this on-line resource I aim to overcome this deficiency, helping to put MCMC in the reach of social scientists. The resource comprises: (*) a set of worked examples (*) data and programs (*) links to other relevant web sites (*) notes and papers At the meetings in Atlanta, I will present two of the worked examples, which are part of this document: (*) Cosponsor: computing auxiliary quantities from MCMC output (e.g., percent correctly predicted in a logit/probit model of legislative behavior; cf Herron 1999). (*) Delegation: estimating a time-series model for ordinal data (e.g., changes to the U.S. president's discretionary power in trade policy, 1890-1990; cf Epstein and O'Halloran 1996).

2
Paper
Democracy as a Latent Variable
Treier, Shawn
Jackman, Simon

Uploaded 07-16-2003
Keywords democracy
Polity
measurement
latent variables
Bayesian statistics
item-response model
ordinal data
latent class analysis
democratic peace
Markov chain Monte Carlo
Abstract Measurement is critical to the social scientific enterprise. Many key concepts in social-scientific theories are not observed directly, and researchers rely on assumptions (tacitly or explicitly, via formal measurement models) to operationalize these concepts in empirical work. In this paper we apply formal, statistical measurement models to the Polity indicators of democracy and autocracy, used widely in studies of international relations. In so doing, we make explicit the hitherto implicit assumptions underlying scales built using the Polity indicators. We discuss two models: one in which democracy is operationalized as a latent continuous variable, and another in which democracy is operationalized as a latent class. Our modeling approaches allow us to assess the measurement error in the resulting measure of democracy. We show that this measurement error is considerable, and has substantive consequences when using a measure of democracy as an independent variable in cross-national statistical analysis. Our analysis suggests that skepticism as to the precision of the Polity democracy scale is well-founded, and that many researchers have been overly sanguine about the properties of the Polity democracy scale in applied statistical work.

3
Paper
Nonparametric Priors For Ordinal Bayesian Social Science Models: Specification and Estimation
Gill, Jeff
Casella, George

Uploaded 08-21-2008
Keywords generalized linear mixed model
ordered probit
Bayesian approaches
nonparametric priors
Dirichlet process mixture models
nonparametric Bayesian inference
Abstract A generalized linear mixed model, ordered probit, is used to estimate levels of stress in presidential political appointees as a means of understanding their surprisingly short tenures. A Bayesian approach is developed, where the random effects are modeled with a Dirichlet process mixture prior, allowing for useful incorporation of prior information, but retaining some vagueness in the form of the prior. Applications of Bayesian models in the social sciences are typically done with ``noninformative'' priors, although some use of informed versions exists. There has been disagreement over this, and our approach may be a step in the direction of satisfying both camps. We give a detailed description of the data, show how to implement the model, and describe some interesting conclusions. The model utilizing a nonparametric prior fits better and reveals more information in the data than standard approaches.

4
Paper
Modeling Electoral Coordination: Voters, Parties and Legislative Lists in Uruguay
Levin, Ines
Katz, Gabriel

Uploaded 04-20-2011
Keywords electoral coordination
number of parties
Bayesian estimation
multilevel modeling
strategic voting
Abstract During each electoral period, the strategic interaction between voters and political elites determines the number of viable candidates in a district. In this paper, we implement a hierarchical seemingly unrelated regression model to explain electoral coordination at the district level in Uruguay as a function of district magnitude, previous electoral outcomes and electoral regime. Elections in this country are particularly useful to test for institutional effects on the coordination process due to the large variations in district magnitude, to the simultaneity of presidential and legislative races held under different rules, and to the reforms implemented during the period under consideration. We find that district magnitude and electoral history heuristics have substantial effects on the number of competing and voted-for parties and lists. Our modeling approach uncovers important interaction-effects between the demand and supply side of the political market that were often overlooked in previous research.

5
Paper
Prior Distributions for Variance Parameters in Hierarchical Models
Gelman, Andrew

Uploaded 03-28-2004
Keywords Bayesian inference
hierarchical model
multilevel model
noninformative prior distribution
weakly informative prior distribution
Abstract Various noninformative prior distributions have been suggested for scale parameters in hierarchical models. We construct a new folded-noncentral-$t$ family of conditionally conjugate priors for hierarchical standard deviation parameters, and then consider noninformative and weakly informative priors in this family. We use an example to illustrate serious problems with the inverse-gamma family of "noninformative" prior distributions. We suggest instead to use a uniform prior on the hierarchical standard deviation, using the half-$t$ family when the number of groups is small and in other settings where a weakly informative prior is desired.

6
Paper
Candidate Viability and Voter Learning in the Presidential Nomination Process
Paolino, Philip

Uploaded 08-30-1999
Keywords beta distribution
maximum likelihood
heterogeneity
Bayesian
Abstract Candidates' viability and momentum are important features of the presidential nomination process in the United States, and much work has examined how both influence the outcome of the nomination campaign (e.g. Aldrich 1980a, Aldrich 1980b, Bartels 1988, Brady and Johnston 1987) Previous treatments, however, have focused upon candidates' expectations of winning or losing the nomination. A critical feature that has been mentioned, but not addressed directly is the volatility of these expectations. In this paper, I use a view of viability and momentum that considers both expectations and the variance of the public's perceptions about candidates' viability which allows us to examine how voters use new information to update their beliefs about both elements of candidates' viability and provides a basis for assessing candidates' potential momentum.

7
Paper
A Bayesian analysis of the multinomial probit model using marginal data augmentation
Imai, Kosuke
van Dyk, David A.

Uploaded 08-21-2002
Keywords Bayesian analysis
Data augmentation
Prior distributions
Probit models
Rate of convergence
Abstract We introduce a set of new Markov chain Monte Carlo algorithms for Bayesian analysis of the multinomial probit model. Our Bayesian representation of the model places a new, and possibly improper, prior distribution directly on the identifiable parameters and thus is relatively easy to interpret and use. Our algorithms, which are based on the method of marginal data augmentation, involve only draws from standard distributions and dominate other available Bayesian methods in that they are as quick to converge as the fastest methods but with a more attractive prior specification.

8
Paper
What Can Be Learned from a Simple Table? Bayesian Inference and Sensitivity Analysis for Causal Effects from 2x2 and 2x2xK Tables in the Presence of Unmeasured Confounding
Quinn, Kevin

Uploaded 09-07-2008
Keywords causal inference
bayesian inference
sensitivity analysis
unmeasured confounding
Abstract What, if anything, should one infer about the causal effect of a binary treatment on a binary outcome from a $2 imes 2$ cross-tabulation of non-experimental data? Many researchers would answer ``nothing'' because of the likelihood of severe bias due to the lack of adjustment for key confounding variables. This paper shows that such a conclusion is unduly pessimistic. Because the complete data likelihood under arbitrary patterns of confounding factorizes in a particularly convenient way, it is possible to parameterize this general situation with four easily interpretable parameters. Subjective beliefs regarding these parameters are easily elicited and subjective statements of uncertainty become possible. This paper also develops a novel graphical display called the confounding plot that quickly and efficiently communicates all patterns of confounding that would leave a particular causal inference relatively unchanged.

9
Paper
Bayesian Measures of Explained Variance and Pooling in Multilevel (Hierarchical) Models
Gelman, Andrew
Pardoe, Iain

Uploaded 04-16-2004
Keywords adjusted R-squared
Bayesian inference
hierarchical model
multilevel regression
partial pooling
shrinkage
Abstract Explained variance (R2) is a familiar summary of the fit of a linear regression and has been generalized in various ways to multilevel (hierarchical) models. The multilevel models we consider in this paper are characterized by hierarchical data structures in which individuals are grouped into units (which themselves might be further grouped into larger units), and there are variables measured on individuals and each grouping unit. The models are based on regression relationships at different levels, with the first level corresponding to the individual data, and subsequent levels corresponding to between-group regressions of individual predictor effects on grouping unit variables. We present an approach to defining R2 at each level of the multilevel model, rather than attempting to create a single summary measure of fit. Our method is based on comparing variances in a single fitted model rather than comparing to a null model. In simple regression, our measure generalizes the classical adjusted R2. We also discuss a related variance comparison to summarize the degree to which estimates at each level of the model are pooled together based on the level-specific regression relationship, rather than estimated separately. This pooling factor is related to the concept of shrinkage in simple hierarchical models. We illustrate the methods on a dataset of radon in houses within counties using a series of models ranging from a simple linear regression model to a multilevel varying-intercept, varying-slope model.

10
Paper
Operationalizing and Testing Spatial Theories of Voting
Quinn, Kevin M.
Martin, Andrew D.

Uploaded 04-15-1998
Keywords spatial voting
factor analysis
multinomial probit
multinomial logit
Bayesian inference
model comparison
Bayes factors
MCMC
Dutch politics
Danish politics
Abstract Spatial models of voting behavior provide the foundation for a substantial number of theoretical results. Nonetheless, empirical work involving the spatial model faces a number of potential difficulties. First, measures of the latent voter and candidate issue positions must be obtained. Second, evaluating the fit of competing statistical models of voter choice is often more complicated than previously realized. In this paper, we discuss precisely these issues. We argue that confirmatory factor analysis applied to mass-level issue preference questions is an attractive means of measuring voter ideal points. We also show how party issue positions can be recovered using a variation of this strategy. We go on to discuss the problems of assessing the fit of competing statistical models (multinomial logit vs. multinomial probit) and competing explanations (those based on spatial theory vs. those derived from other theories of voting such as sociological theories). We demonstrate how the Bayesian perspective not only provides computational advantages in the case of fitting the multinomial probit model, but also how it facilitates both types of comparison mentioned above. Results from the Netherlands and Denmark suggest that even when the computational cost of multinomial probit is disregarded, the decision whether to use multinomial probit (MNP) or multinomial logit (MNL) is not clear-cut.

11
Paper
Estimating Incumbency Advantage and Campaign Spending Effect without the Simultaneity Bias
Fukumoto, Kentaro

Uploaded 07-16-2006
Keywords Incumbency Advantage
Campaign Spending
Simultaneity Bias
Bayesian Nash equilibria
normal vote
Abstract In estimating incumbency advantage and campaign spending effect, simultaneity problem is composed of stochastic dependence and parametric dependence. Scholars have tried to solve the former, while the present paper intends to tackle the latter. Its core idea is to estimate parameters by maximizing likelihood of all endogenous variables (vote, both parties' candidate qualities and campaign spending) simultaneously. In order to do it, I take advantage of theories of electoral politics rigorously, model each endogenous variables by the others (or their expectation), derive Bayesian Nash equilibria, and plug them into my estimator. I show superiority of my model compared to the conventional estimators by Monte Carlo simulation. Empirical application of this model to the recent U.S. House election data demonstrates that incumbency advantage is smaller than previously shown and that entry of incumbent and strong challenger is motivated by electoral prospect.

12
Paper
Bayesian Combination of State Polls and Election Forecasts
Lock, Kari
Gelman, Andrew

Uploaded 09-21-2008
Keywords election prediction
pre-election polls
Bayesian updating
shrinkage estimation
Abstract In February of 2008, SurveyUSA polled 600 people in each state and asked who they would vote for in either head-to-head match-up: Obama vs. McCain, and Clinton vs. McCain. Here we integrate these polls with prior information; how each state voted in comparison to the national outcome in the 2004 election. We use Bayesian methods to merge prior and poll data, weighting each by its respective information. The variance for our poll data incorporates both sampling variability and variability due to time before the election, estimated using pre-election poll data from the 2000 and 2004 elections. The variance for our prior data is estimated using the results of the past nine presidential elections. The union of prior and poll data results in a posterior distribution predicting how each state will vote, in turn giving us posterior intervals for both the popular and electoral vote outcomes of the 2008 presidential election. Lastly, these posterior distributions are updated with the most recent poll data as of August, 2008.

13
Paper
Practical Issues in Implementing and Understanding Bayesian Ideal Point Estimation
Bafumi, Joseph
Gelman, Andrew
Park, David K.
Kaplan, Noah

Uploaded 06-11-2004
Keywords Ideal points
Bayesian
Logistic regression
Rasch model
Abstract In recent years, logistic regression (Rasch) models have been used in political science for estimating ideal points of legislators and Supreme Court justices. These models present estimation and identifiability challenges, such as improper variance estimates, scale and translation invariance, reflection invariance, and issues with outliers. We resolve these issues using Bayesian hierarchical modeling, linear transformations, informative regression predictors, and explicit modeling for outliers. In addition, we explore new ways to usefully display inferences and check model fit.

14
Paper
A Random Effects Approach to Legislative Ideal Point Estimation
Bailey, Michael

Uploaded 04-21-1998
Keywords ideal points
random effects models
Bayesian estimation
em algorithm
Abstract Conventionally, scholars use either standard probit/logit techniques or fixed-effect methods to estimate legislative ideal points. However, these methods are unsatisfactory when a limited number of votes are available: standard probit/logit methods are poorly equipped to handle multiple votes and fixed-effect models disregard serious ``incidental parameter'' problems. In this paper I present an alternative approach that moves beyond single-vote probit/logit analysis without requiring the large number of votes needed for fixed-effects models. The method is based on a random effects, panel logit framework that models ideal points as stochastic functions of legislator characteristics. Monte Carlo results and an application to trade politics demonstrate the practical usefulness of the method.

15
Paper
An Automated Method of Topic-Coding Legislative Speech Over Time with Application to the 105th-108th U.S. Senate
Quinn, Kevin
Monroe, Burt
Colaresi, Michael
Crespin, Michael
Radev, Dragomir

Uploaded 07-18-2006
Keywords legislatures
agendas
content analysis
Bayesian
time series
cluster analysis
unsupervised learning
Abstract We describe a method for statistical learning from speech documents that we apply to the Congressional Record in order to gain new insight into the dynamics of the political agenda. Prior efforts to evaluate the attention of elected representatives across topic areas have largely been expensive manual coding exercises and are generally circumscribed along one or more features of detail: limited time periods, high levels of temporal aggregation, and coarse topical categories. Conversely, the Congressional Record has scarcely been used for such analyses, largely because it contains too much information to absorb manually. We describe here a method for inferring, through the patterns of word choice in each speech and the dynamics of word choice patterns across time, (a) what the topics of speeches are, and (b) the probability that attention will be paid to any given topic or set of topics over time. We use the model to examine the agenda in the United States Senate from 1997-2004, based on a new database of over 70 thousand speech documents containing over 70 million words. We estimate the model for 42 topics and provide evidence that we can reveal speech topics that are both distinctive and inter-related in substantively meaningful ways. We demonstrate further that the dynamics our model gives us leverage into important questions about the dynamics of the political agenda.

16
Paper
Prior distributions for Bayesian data analysis in political science
Gelman, Andrew

Uploaded 02-25-2009
Keywords Bayesian inference
hierarchical models
mixture models
prior information
Abstract Prior information is often what makes Bayesian inference work. In the political science examples of which we are aware aware, information needs to come in, whether as regression predictors or regularization (that is, prior distributions) on parameters. We illustrate with a few examples from our own research.

17
Paper
Estimating incumbency advantage and its variation, as an example of a before/after study
Gelman, Andrew
Huang, Zaiying

Uploaded 02-07-2003
Keywords Bayesian inference
before-after study
Congressional elections
Gibbs
Abstract Incumbency advantage is one of the most studied features in American legislative elections. In this paper, we construct and implement an estimate that allows incumbency advantage to vary between individual incumbents. This model predicts that open-seat elections will be less variable than those with incumbents running, an observed empirical pattern that is not explained by previous models. We apply our method to the U.S. House of Representatives in the twentieth century: our estimate of the overall pattern of incumbency advantage over time is similar to previous estimates (although slightly lower), and we also find a pattern of increasing variation. In addition to the application to incumbency advantage, our approach represents a new method, using multilevel modeling, for estimating effects in before/after studies.

18
Paper
Rational Expectations Coordinating Voting in American Presidential and House Elections
Mebane, Walter R.

Uploaded 07-08-1998
Keywords coordinating voting
probabilistic voting
spatial voting
retrospective voting
policy moderation
presidential elections
congressional elections
ticket splitting
rational expectations
voter equilibrium
Bayesian-Nash equilibrium
generalized extreme value model
nonparametric
Monte Carlo integration
maximum likelihood
Abstract I define a probabilistic model of individuals' presidential-year vote choices for President and for the House of Representatives in which there is a coordinating (Bayesian Nash) equilibrium among voters based on rational expectations each voter has about the election outcomes. I estimate the model using data from the six American National Election Study Pre-/Post-Election Surveys of years 1976--1996. The coordinating model passes a variety of tests, including a test against a majoritarian model in which there is rational ticket splitting but no coordination. The results give strong individual-level support to Alesina and Rosenthal's theory that voters balance institutions in order to moderate policy. The estimates describe vote choices that strongly emphasize the presidential candidates. I also find that a voter who says economic conditions have improved puts more weight on a discrepancy between the voter's ideal point and government policy with a Democratic President than on a discrepancy of the same size with a Republican President.

19
Paper
Presidential Approval: the case of George W. Bush
Beck, Nathaniel
Jackman, Simon
Rosenthal, Howard

Uploaded 07-19-2006
Keywords presidential approval
public opinion
polls
house effects
dynamic linear model
Bayesian statistics
Markov chain Monte Carlo
state space
pages of killer graphs
Abstract We use a Bayesian dynamic linear model to track approval for George W. Bush over time. Our analysis deals with several issues that have been usually addressed separately in the extant literature. First, our analysis uses polling data collected at a higher frequency than is typical, using over 1,100 published national polls, and data on macro-economic conditions collected at the weekly level. By combining this much poll information, we are much better poised to examine the public's reactions to events over shorter time scales than can the typical analysis of approval that utilizes monthly or quarterly approval. Second, our statistical modeling explicitly deals with the sampling error of these polls, as well as the possibility of bias in the polls due to house effects. Indeed, quite aside from the question of ``what drives approval?'', there is considerable interest in the extent to which polling organizations systematically diverge from one another in assessing approval for the president. These bias parameters are not only necessary parts of any realistic model of approval that utilizes data from multiple polling organizations, but easily estimated via the Bayesian dynamics linear model.

20
Paper
Spike and Slab Prior Distributions for Simultaneous Bayesian Hypothesis Testing, Model Selection, and Prediction, of Nonlinear Outcomes
Pang, Xun
Gill, Jeff

Uploaded 07-13-2009
Keywords Spike and Slab Prior
Hypothesis Testing
Bayesian Model Selection
Bayesian Model Averaging
Adaptive Rejection Sampling
Generalized Linear Model
Abstract A small body of literature has used the spike and slab prior specification for model selection with strictly linear outcomes. In this setup a two-component mixture distribution is stipulated for coefficients of interest with one part centered at zero with very high precision (the spike) and the other as a distribution diffusely centered at the research hypothesis (the slab). With the selective shrinkage, this setup incorporates the zero coefficient contingency directly into the modeling process to produce posterior probabilities for hypothesized outcomes. We extend the model to qualitative responses by designing a hierarchy of forms over both the parameter and model spaces to achieve variable selection, model averaging, and individual coefficient hypothesis testing. To overcome the technical challenges in estimating the marginal posterior distributions possibly with a dramatic ratio of density heights of the spike to the slab, we develop a hybrid Gibbs sampling algorithm using an adaptive rejection approach for various discrete outcome models, including dichotomous, polychotomous, and count responses. The performance of the models and methods are assessed with both Monte Carlo experiments and empirical applications in political science.

21
Paper
Taking the State Space Seriously: The Dynamic Linear Model and Bayesian Time Series Analysis
Buckley, Jack

Uploaded 08-02-2002
Keywords time series
bayesian
state space
think tanks
business
Abstract No abstract submitted.

22
Paper
The Problem with Quantitative Studies of International Conflict
Beck, Nathaniel
King, Gary
Zeng, Langche

Uploaded 07-15-1998
Keywords Conflict
logit
neural networks
forecasting
Bayesian analysis
Abstract Despite immense data collections, prestigious journals, and sophisticated analyses, empirical findings in the literature on international conflict are frequently unsatisfying. Statistical results appear to change from article to article and specification to specification. Very few relationships hold up to replication with even minor respecification. Accurate forecasts are nonexistent. We provide a simple conjecture about what accounts for this problem, and offer a statistical framework that better matches the substantive issues and types of data in this field. Our model, a version of a ``neural network'' model, forecasts substantially better than any previous effort, and appears to uncover some structural features of international conflict.

23
Paper
Expressive Bayesian Voters, their Turnout Decisions, and Double Probit
Achen, Christopher

Uploaded 07-17-2006
Keywords turnout
expressive
Bayesian
probit
scobit
EITM
Abstract Voting is an expressive act. Since people are not born wanting to express themselves politically, the desire to vote must be acquired, either by learning about the candidates, by using party identification as a cognitive shortcut, or by contact from a trusted source. Modeled as Bayesian updating, this simple explanatory framework has dramatic implications for the understanding of voter turnout. It mathematically implies the main empirical generalizations familiar from the literature, it predicts hitherto unnoticed patterns that appear in turnout data, it provides a better fitting statistical model (double probit) for sample surveys of turnout, and it allows researchers to forecast turnout patterns in new elections when circumstances change. Thus the case is strengthened for the Bayesian voter model as a central organizing principle for public opinion and voting behavior.

24
Paper
Joint Modeling of Dynamic and Cross-Sectional Heterogeneity: Introducing Hidden Markov Panel Models
Park, Jong Hee

Uploaded 07-14-2009
Keywords Bayesian statistics
Fixed-effects
Hidden Markov models
Markov chain Monte Carlo methods
Random-effects
Reversible jump Markov chain Monte Carlo
Abstract Researchers working with panel data sets often face situations where changes in unobserved factors have produced changes in the cross-sectional heterogeneity across time periods. Unfortunately, conventional statistical methods for panel data are based on the assumption that the unobserved cross-sectional heterogeneity is time constant. In this paper, I introduce statistical methods to diagnose and model changes in the unobserved heterogeneity. First, I develop three combinations of a hidden Markov model with panel data models using the Bayesian framework; (1) a baseline hidden Markov panel model with varying fixed effects and varying random effects; (2) a hidden Markov panel model with varying fixed effects; and (3) a hidden Markov panel model with varying intercepts. Second, I present model selection methods to diagnose the dynamic heterogeneity using the marginal likelihood method and the reversible jump Markov chain Monte Carlo method. I illustrate the utility of these methods using two important ongoing political economy debates; the relationship between income inequality and economic growth and the effect of institutions on income inequality.

25
Paper
Unresponsive, Unpersuaded: The Unintended Consequences of Voter Persuasion Efforts
Bailey, Michael
Hopkins, Daniel
Rogers, Todd

Uploaded 08-09-2013
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 follow-up survey, lowering the response rate sharply among infrequent voters. Second, various statistical methods to address the resulting biases converge on a counter-intuitive 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.

26
Paper
Connecting Interest Groups and Congress: A New Approach to Understanding Interest Group Success
Victor, Jennifer Nicoll

Uploaded 07-16-2002
Keywords Interest Groups
Congress
Multiple Imputation
Bayesian Information Criterion
Ordinal Probit
Non-nested Models
Legislative Context
Abstract The primary challenge in explaining interest group legislative success in Congress has been methodological. The discipline requires at least two critical elements to make progress on this important question. First, we need a theory that accounts for the highly interactive spatial game between interest groups and legislators. Second, the discipline needs an empirical model that associates interest groups and their activities with specific congressional bills. In this project I begin to contribute to our understanding of the complex relationship between interest groups and Congress. I develop a theory of group success that is based upon the strategies in which groups engage, the groups' organizational capacity, and the strategic context of legislation. I predict that groups will tailor their activities (and strategically spend their resources) in Congress based upon two critical factors: whether the group supports or opposes the legislation, and the legislative environment for the bill. To test this model I develop a unique sampling procedure and survey design. I use legislative hearings to generate a sample of groups that are associated with specific issues and survey them about their activities on those issues. Then, I associate each group's issue with a specific bill in Congress. I then track the bill to discern its final status. I create a dependent variable of interest group success that is based on the group's position (favor or oppose) and the final status of the bill. This sampling procedure and dependent variable allow me to make inferences about group behavior over specific legislative proposals. I develop independent variables of group activity, group organizational capacity, and legislative context from the survey instrument and objective information about the bills. To fill in gaps in the survey data set, I use a multiple imputation method that generates plausible values based on given distributions of data. I estimate two models-one for groups in favor of legislation, and one for opposition groups. The ordinal probit models generally support the theoretical expectations. In sum, I find that groups can best expend their resources in pursuit of rules that advantage their position rather than fighting for bill content.

27
Paper
Pooling Disparate Observations
Bartels, Larry M.

Uploaded 01-01-1995
Keywords induction
statistical inference
Bayesian statistics
econometrics
observations
F-test
pooling
fractional pooling
Abstract Data analysts frequently face difficult choices about whether to pool disparate observations in their statistical analyses. I explore the inferential ramifications of such choices, and propose a new technique, dubbed "fractional pooling," which provides a simple way to incorporate prior beliefs about the theoretical relevance of disparate observations. The technique is easy to implement and has a plausible rationale in Bayesian statistical theory. I illustrate the potential utility of fractional pooling by applying the technique to political data originally analyzed by Ashenfelter (1994), Powell (1982), and Alesina et al. (1993). These examples demonstrate that conventional approaches to analyzing disparate observations can be seriously misleading, and that the approach proposed here can enrich our understanding of the inferential implications of unavoidably subjective judgments about the theoretical relevance of available data.

28
Paper
Designing and Analyzing Randomized Experiments
Horiuchi, Yusaku
Imai, Kosuke
Taniguchi, Naoko

Uploaded 07-05-2005
Keywords Bayesian inference
causal inference
noncompliance
nonresponse
randomized block design
Abstract In this paper, we demonstrate how to effectively design and analyze randomized experiments, which are becoming increasingly common in political science research. Randomized experiments provide researchers with an opportunity to obtain unbiased estimates of causal effects because the randomization of treatment guarantees that the treatment and control groups are on average equal in both observed and unobserved characteristics. Even in randomized experiments, however, complications can arise. In political science experiments, researchers often cannot force subjects to comply with treatment assignment or to provide the information necessary for the estimation of causal effects. Building on the recent statistical literature, we show how to make statistical adjustments for these noncompliance and nonresponse problems when analyzing randomized experiments. We also demonstrate how to design randomized experiments so that the potential impact of such complications is minimized.

29
Paper
Balancing Competing Demands: Position-Taking and Election Proximity in the European Parliament
Lindstaedt, Rene
Slapin, Jonathan
Vander Wielen, Ryan

Uploaded 07-31-2009
Keywords Legislative Politics
European Parliament
Comparative Politics
Bayesian IRT
Parties
Formal Theory
Abstract Parties value unity, yet, members of parliament face competing demands, giving them incentives to deviate from the party. For members of the European Parliament (MEPs), these competing demands are national party and European party group pressures. Here, we look at how MEPs respond to those competing demands. We examine ideological shifts within a single parliamentary term to assess how European Parliament (EP) election proximity affects party group cohesion. Our formal model of legislative behavior with multiple principals yields the following hypothesis: When EP elections are proximate, national party delegations shift toward national party positions, thus weakening EP party group cohesion. For our empirical test, we analyze roll call data from the fifth EP (1999-2004) using Bayesian item response models. We find significant movement among national party delegations as EP elections approach, which is consistent with our theoretical model, but surprising given the existing literature on EP elections as second-order contests.

30
Paper
State-Level Opinions from National Surveys: Poststratification using Hierarchical Logistic Regression
Park, David K.
Gelman, Andrew
Bafumi, Joseph

Uploaded 07-12-2002
Keywords Bayesian Inference
Hierarchical
Logit
Poststratification
Public Opinion
States
Elections
Abstract Previous researchers have pooled national surveys in order to construct state-level opinions. However, in order to overcome the small n problem for less populous states, they have aggregated a decade or more of national surveys to construct their measures. For example, Erikson, Wright and McIver (1993) pooled 122 national surveys conducted over 13 years to produce state-level partisan and ideology estimates. Brace, Sims-Butler, Arceneaux, and Johnson (2002) pooled 22 surveys over a 25-year period to produce state-level opinions on a number of specific issues. We construct a hierarchical logistic regression model for the mean of a binary response variable conditional on poststratification cells. This approach combines the modeling approach often used in small-area estimation with the population information used in poststratification (see Gelman and Little 1997). We produce state-level estimates pooling seven national surveys conducted over a nine-day period. We first apply the method to a set of U.S pre-election polls, poststratified by state, region, as well as the usual demographic variables and evaluate the model by comparing it to state-level election outcomes. We then produce state-level partisan and ideology estimates by comparing it to Erikson, Wright and McIver's estimates.

31
Paper
A Bayesian Method for the Analysis of Dyadic Crisis Data
Smith, Alastair

Uploaded 11-04-1996
Keywords Bayesian model testing
Censored data
Crisis data
Gibbs sampling
Markov chain Monte Carlo
Ordered discrete choice model
Strategic choice
Abstract his paper examines the level of force that nations use during disputes. Suppose that two nations, A and B, are involved in a dispute. Each nation chooses the level of violence that it is prepared to use in order to achieve its objectives. Since there are two opponents making decisions, the outcome of the crisis is determined by a bivariate rather than univariate process. I propose a bivariate ordered discrete choice model to examine the relationship between nation A's decision to use force, nation B's decision to use force, and a series of explanatory variables. The model is estimated in the Bayesian context using a Markov chain Monte Carlo simulation technique. I analyze Bueno de Mesquita and Lalman's (1992) dyadically coded version of the Militarized Interstate Dispute data (Gochman and Moaz 1984). Various models are compared using Bayes Factors. The results indicate that nation A's and nation B's decisions to use force can not be regarded as independent. Bayesian model comparison show that variables derived from Bueno de Mesquita's expected utility theory (1982, 1985; Bueno de Mesquita and Lalman 1986, 1992) provide the best explanatory variables for decision making in crises.

32
Paper
Bayesian and Likelihood Inference for 2 x 2 Ecological Tables: An Incomplete Data Approach
Imai, Kosuke
Lu, Ying
Strauss, Aaron

Uploaded 12-16-2006
Keywords Coarse data
Contextual effects
Data augmentation
EM algorithm
Missing information principle
Nonparametric Bayesian Modeling.
Abstract Ecological inference is a statistical problem where aggregate-level data are used to make inferences about individual-level behavior. Recent years have witnessed resurgent interest in ecological inference among political methodologists and statisticians. In this paper, we conduct a theoretical and empirical study of Bayesian and likelihood inference for 2 x 2 ecological tables by applying the general statistical framework of incomplete data. We first show that the ecological inference problem can be decomposed into three factors: distributional effects which address the possible misspecification of parametric modeling assumptions about the unknown distribution of missing data, contextual effects which represent the possible correlation between missing data and observed variables, and aggregation effects which are directly related to the loss of information caused by data aggregation. We then examine how these three factors affect inference and offer new statistical methods to address each of them. To deal with distributional effects, we propose a nonparametric Bayesian model based on a Dirichlet process prior which relaxes common parametric assumptions. We also specify the statistical adjustments necessary to account for contextual effects. Finally, while little can be done to cope with aggregation effects, we offer a method to quantify the magnitude of such effects in order to formally assess its severity. We use simulated and real data sets to empirically investigate the consequences of these three factors and to evaluate the performance of our proposed methods. C code, along with an easy-to-use R interface, is publicly available for implementing our proposed methods.

33
Paper
Bayesian statistical decision theory and a critical test for substantive significance
Esarey, Justin

Uploaded 09-09-2009
Keywords inference
t-test
substantive significance
Bayesian
Abstract I introduce a new critical test statistic, c*, that uses Bayesian statistical decision theory to help an analyst determine whether quantitative evidence supports the existence of a substantively meaningful relationship. Bayesian statistical decision theory takes a rational choice perspective toward evidence, allowing researchers to ask whether it makes sense to believe in the existence of a statistical relationship given how they value the consequences of correct and incorrect decisions. If a relationship of size c* is not important enough to influence future research and policy advice, then the evidence does not support the existence of a substantively significant effect. A replication of findings from the American Journal of Political Science and Journal of Politics illustrates that statistical significance at conventional levels is neither necessary nor sufficient to accept a hypothesis of substantive significance using c*. I also make software packages available for Stata and R that allow political scientists to easily use c* for inference in their own research.

34
Paper
Moving Mountains: Bayesian Forecasting As Policy Evaluation
Brandt, Patrick T.
Freeman, John R.

Uploaded 04-24-2002
Keywords Bayesian vector autoregression
VAR
policy evaluation
conditional forecasting
Abstract Many policy analysts fail to appreciate the dynamic, complex causal nature of political processes. We advocate a vector autoregression (VAR) based approach to policy analysis that accounts for various multivariate and dynamic elements in policy formulation and for both dynamic and specification uncertainty of parameters. The model we present is based on recent developments in Bayesian VAR modeling and forecasting. We present an example based on work in Goldstein et al. (2001) that illustrates how a full accounting of the dynamics and uncertainty in multivariate data can lead to more precise and instructive results about international mediation in Middle Eastern conflict.

35
Paper
Not Asked and Not Answered: Multiple Imputation for Multiple Surveys
Gelman, Andrew
King, Gary
Liu, Chuanhai

Uploaded 10-27-1997
Keywords Bayesian inference
cluster sampling
diagnostics
hierarchical models
ignorable nonresponse
missing data
political science
sample surveys
stratified sampling
multiple imputation
Abstract We present a method of analyzing a series of independent cross-sectional surveys in which some questions are not answered in some surveys and some respondents do not answer some of the questions posed. The method is also applicable to a single survey in which different questions are asked, or different sampling methods used, in different strata or clusters. Our method involves multiply-imputing the missing items and questions by adding to existing methods of imputation designed for single surveys a hierarchical regression model that allows covariates at the individual and survey levels. Information from survey weights is exploited by including in the analysis the variables on which the weights were based, and then reweighting individual responses (observed and imputed) to estimate population quantities. We also develop diagnostics for checking the fit of the imputation model based on comparing imputed to non-imputed data. We illustrate with the example that motivated this project --- a study of pre-election public opinion polls, in which not all the questions of interest are asked in all the surveys, so that it is infeasible to impute each survey separately.

36
Paper
Modeling Foreign Direct Investment as a Longitudinal Social Network
Jensen, Nathan
Martin, Andrew
Westveld, Anton

Uploaded 07-13-2007
Keywords foreign direct investment
social network data
longitudinal data
hierarchical modeling
mixture modeling
Bayesian inference.
Abstract An extensive literature in international and comparative political economy has focused on the how the mobility of capital affects the ability of governments to tax and regulate firms. The conventional wisdom holds that governments are in competition with each other to attract foreign direct investment (FDI). Nation-states observe the fiscal and regulatory decisions of competitor governments, and are forced to either respond with policy changes or risk losing foreign direct investment, along with the politically salient jobs that come with these investments. The political economy of FDI suggests a network of investments with complicated dependencies. We propose an empirical strategy for modeling investment patterns in 24 advanced industrialized countries from 1985-2000. Using bilateral FDI data we estimate how increases in flows of FDI affect the flows of FDI in other countries. Our statistical model is based on the methodology developed by Westveld & Hoff (2007). The model allows the temporal examination of each notion's activity level in investing, attractiveness to investors, and reciprocity between pairs of nations. We extend the model by treating the reported inflow and outflow data as independent replicates of the true value and allowing for a mixture model for the fixed effects portion of the network model. Using a fully Bayesian approach, we also impute missing data within the MCMC algorithm used to fit the model.

37
Paper
Characterizing the variance improvement in linear Dirichlet random effects models
Kyung, Minjung
Gill, Jeff
Casella, George

Uploaded 09-11-2009
Keywords Dirichlet processes
mixture models
Bayesian nonparametrics
Abstract An alternative to the classical mixed model with normal random effects is to use a Dirichlet process to model the random effects. Such models have proven useful in practice, and we have observed a noticeable variance reduction, in the estimation of the fixed effects, when the Dirichlet process is used instead of the normal. In this paper we formalize this notion, and give a theoretical justification for the expected variance reduction. We show that for almost all data vectors, the posterior variance from the Dirichlet random effects model is smaller than that form the normal random effects model. Forthcoming: Statistics and Probability Letters

38
Paper
Did Illegally Counted Overseas Absentee Ballots Decide the 2000 U.S. Presidential Election?
Imai, Kosuke
King, Gary

Uploaded 02-13-2002
Keywords 2000 U.S. Presidential Election
Ecological Inference
Bayesian Model Averaging
Abstract Although not widely known until much later, Al Gore received 202 more votes than George W. Bush on election day in Florida. George W. Bush is president because he overcame his election day deficit with overseas absentee ballots that arrived and were counted after election day. In the final official tally, Bush received 537 more votes than Gore. These numbers are taken from the official results released by the Florida Secretary of State's office and so do not reflect overvotes, undervotes, unsuccessful litigation, butterfly ballot problems, recounts that might have been allowed but were not, or any other hypothetical divergence between voter preferences and counted votes. After the election, the New York Times conducted a six month long investigation and found that 680 of the overseas absentee ballots were illegally counted, and no partisan, pundit, or academic has publicly disagreed with their assessment. In this paper, we describe the statistical procedures we developed and implemented for the Times to ascertain whether disqualifying these 680 ballots would have changed the outcome of the election. The methods involve adding formal Bayesian model averaging procedures to King's (1997) ecological inference model. Formal Bayesian model averaging has not been used in political science but is especially useful when substantive conclusions depend heavily on apparently minor but indefensible model choices, when model generalization is not feasible, and when potential critics are more partisan than academic. We show how we derived the results for the Times so that other scholars can use these methods to make ecological inferences for other purposes. We also present a variety of new empirical results that delineate the precise conditions under which Al Gore would have been elected president, and offer new evidence of the striking effectiveness of the Republican effort to convince local election officials to count invalid ballots in Bush counties and not count them in Gore counties.

39
Paper
Testing Theories Involving Strategic Choice: The Example of Crisis Escalation
Smith, Alastair

Uploaded 07-23-1997
Keywords Strategic choice
Bayesian model testing
Markov chain Monte Carlo simulation
multi-variate probit
crisis escalation
war
Abstract If we believe that politics involves a significant amount of strategic interaction then classical statistical tests, such as Ordinary Least Squares, Probit or Logit, cannot give us the right answers. This is true for two reasons: The dependent variables under observation are interdependent-- that is the essence of game theoretic logic-- and the data is censored -- that is an inherent feature of off the path expectations that leads to selection effects. I explore the consequences of strategic decision making on empirical estimation in the context of international crisis escalation. I show how and why classical estimation techniques fail in strategic settings. I develop a simple strategic model of decision making during crises. I ask what this explanation implies about the distribution of the dependent variable: the level of violence used by each nation. Counterfactuals play a key role in this theoretical explanation. Yet, conventional econometric techniques take no account of unrealized opportunities. For example, suppose a weak nation (B) is threatened by a powerful neighbor (A). If we believe that power strongly influences the use of force then the weak nation realizes that the aggressor's threats are probably credible. Not wishing to fight a more powerful opponent, nation B is likely to acquiesce to the aggressor's demands. Empirically, we observe A threaten B. The actual level of violence that A uses is low. However, the theoretical model suggests that B acquiesced precisely because A would use force. Although the theoretical model assumes a strong relationship between strength and the use of force, traditional techniques find a much weaker relationship. Our ability to observe whether nation A is actually prepared to use force is censored when nation B acquiesces. I develop a Strategically Censored Discrete Choice (SCDC) model which accounts for the interdependent and censored nature of strategic decision making. I use this model to test existing theories of dispute escalation. Specifically, I analyze Bueno de Mesquita and Lalman's (1992) dyadically coded version of the Militarized Interstate Dispute data (Gochman and Moaz 1984). I estimate this model using a Bayesian Markov chain Monte Carlo simulation method. Using Bayesian model testing, I compare the explanatory power of a variety of theories. I conclude that strategic choice explanations of crisis escalation far out-perform non-strategic ones.

40
Paper
Bayesian Analysis of Structural Changes: Historical Changes in US Presidential Uses of Force Abroad
Park, Jong Hee

Uploaded 07-16-2007
Keywords structural changes
changepoint models
discrete time series data
use of force data
state space models
time-varying parameter models
Bayesian inference
Abstract While many theoretical models in political science are inspired by structural changes in politics, most empirical methods assume stable patterns of causal processes and fail to capture dynamic changes in theoretical relationships. In this paper, I introduce an efficient Bayesian approach to the multiple changepoint problem presented by Chib (1998) and discuss the utility of the Bayesian changepoint models in the context of generalized linear models. As an illustration, I revisit the debate over whether and how U.S. presidents have used forces abroad in response to domestic factors since 1890.

41
Paper
Penalized Regression, Standard Errors, and Bayesian Lassos
Kyung, Minjung
Gill, Jeff
Ghosh, Malay
Casella, George

Uploaded 02-23-2010
Keywords model selection
lassos
Bayesian hierarchical models
LARS algorithm
EM/Gibbs sampler
Geometric Ergodicity
Gibbs Sampling
Abstract Penalized regression methods for simultaneous variable selection and coefficient estimation, especially those based on the lasso of Tibshirani (1996), have received a great deal of attention in recent years, mostly through frequentist models. Properties such as consistency have been studied, and are achieved by different lasso variations. Here we look at a fully Bayesian formulation of the problem, which is flexible enough to encompass most versions of the lasso that have been previously considered. The advantages of the hierarchical Bayesian formulations are many. In addition to the usual ease-of-interpretation of hierarchical models, the Bayesian formulation produces valid standard errors (which can be problematic for the frequentist lasso), and is based on a geometrically ergodic Markov chain. We compare the performance of the Bayesian lassos to their frequentist counterparts using simulations and data sets that previous lasso papers have used, and see that in terms of prediction mean squared error, the Bayesian lasso performance is similar to and, in some cases, better than, the frequentist lasso.

42
Paper
Random Coefficient Models for Time-Series--Cross-Section Data: The 2001 Version
Beck, Nathaniel
Katz, Jonathan

Uploaded 07-17-2001
Keywords random coefficients
generalized least squares
empirical Bayesian
Stein-rule
TCSC
Abstract This paper considers random coefficient models (RCMs) for time-series--cross-section data. These models allow for unit to unit variation in the model parameters. After laying out the various models, we assess several issues in specifying RCMs. We then consider the finite sample properties of some standard RCM estimators, and show that the most common one, associated with Hsaio, has very poor properties. These analyses also show that a somewhat awkward combination of estimators based on Swamy's work performs reasonably well; this awkward estimator and a Bayes estimator with an uninformative prior (due to Smith) seem to perform best. But we also see that estimators which assume full pooling perform well unless there is a large degree of unit to unit parameter heterogeneity. We also argue that the various data driven methods (whether classical or empirical Bayes or Bayes with gentle priors) tends to lead to much more heterogeneity than most political scientists would like. We speculate that fully Bayesian models, with a variety of informative priors, may be the best way to approach RCMs.

43
Paper
Too many Variables? A Comment on Bartels' ModelAveraging Proposal
Erikson, Robert S.
Wright, Gerald C.
McIver, John P.

Uploaded 07-18-1997
Keywords Bayes Factor
Bayesian Information Criterion
Bayesian statistics
model averaging
model specification
specification uncertainty
Bartels
Abstract Abstract: Bartels (1997) popularizes the procedure of model- averaging (Raftery, 1995, 1997), making some important innovations of his own along the way. He offers his methodology as a technology for exposing excessive specification searches in other peoples' research. As a demonstration project, Bartels applied his version of model- averaging to a portion of our work on state policy and purports to detect evidence of considerable model uncertainty. . In response, we argue that Bartels' extensions of model averaging methodology are ill-advised, and show that our challenged findings hold up under the scrutiny of the original Raftery-type model averaging.

44
Paper
The Spatial Probit Model of Interdependent Binary Outcomes: Estimation, Interpretation, and Presentation
Franzese, Robert
Hays, Jude

Uploaded 07-20-2007
Keywords Spatial Probit
Bayesian Gibbs-Sampler Estimator
Recursive Importance-Sampling Estimator
Interdependence
Diffusion
Contagion
Emulation
Abstract We have argued and shown elsewhere the ubiquity and prominence of spatial interdependence in political science research and noted that much previous practice has neglected this interdependence or treated it solely as nuisance to the serious detriment of sound inference. Previously, we considered only linear-regression models of spatial and/or spatio-temporal interdependence. In this paper, we turn to binary-outcome models. We start by stressing the ubiquity and centrality of interdependence in binary outcomes of interest to political and social scientists and note that, again, this interdependence has been ignored in most contexts where it likely arises and that, in the few contexts where it has been acknowledged, the endogeneity of the spatial lag has not be recognized. Next, we explain some of the severe challenges for empirical analysis posed by spatial interdependence in binary-outcome models, and then we follow recent advances in the spatial-econometric literature to suggest Bayesian or recursive-importance-sampling (RIS) approaches for tackling estimation. In brief and in general, the estimation complications arise because among the RHS variables is an endogenous weighted spatial-lag of the unobserved latent outcome, y*, in the other units; Bayesian or RIS techniques facilitate the complicated nested optimization exercise that follows from that fact. We also advance that literature by showing how to calculate estimated spatial effects (as opposed to parameter estimates) in such models, how to construct confidence regions for those (adopting a simulation strategy for the purpose), and how to present such estimates effectively.

45
Paper
No News is News: Non-Ignorable Non-Response in Roll-Call Data Analysis
Rosas, Guillermo
Shomer, Yael
Haptonstahl, Stephen

Uploaded 07-10-2010
Keywords rollcall
voting
abstention
missing
Bayesian
IRT
Abstract Roll-call votes are widely employed to infer the ideological proclivities of legislators, even though inferences based on roll-call data are accurate reflections of underlying policy preferences only under stringent assumptions. We explore the consequences of violating one such assumption, namely, the ignorability of the process that generates non-response in roll calls. We offer a reminder of the inferential consequences of ignoring certain processes of non-response, a basic estimation framework to model non-response and vote choice concurrently, and models for two theoretically relevant processes of non-ignorable missingness. We reconsider the "most liberal Senator" question that comes up during election times every four years in light of our arguments and show how we inferences about ideal points can improve by incorporating a priori information about the process that generates abstentions.

46
Paper
Pre-Election Polls in Nation and State: A Dynamic Bayesian Hierarchical Model
Franklin, Charles

Uploaded 07-17-2001
Keywords campaigns
polling
aggregation
Bayesian
hierarchical models
Abstract A vast number of national trial heat polls are conducted in the months preceding a presidential election. But as was dramatically demonstrated in 2000, candidates must win states to win the presidency, not just win popular votes. The density of state level polling is much less than that for the nation as a whole. This makes efforts to track candidate support at the state level, and to estimate campaign effects in the states, very difficult. This paper develops a Bayesian hierarchical model of trial heat polls which uses state and national polling data, plus measures of campaign effort in each state, to estimate candidate support between observed state polls. At a technical level, the Bayesian approach provides not only estimates of support but also easily understood estimates of the uncertainty of those estimates. At an applied level, this method can allow campaigns to target polling in states that are most likely to be changing while being alerted to potential shifts in states that are not as frequently polled.

47
Paper
Recent Developments in Econometric Modelling: A Personal Viewpoint
Maddala, G.S.

Uploaded 07-17-1997
Keywords dynamic panel data models
dynamic models with limited dependent variables
unit roots
cointegration
VAR's
Bayesian
Abstract The quotation above (more than three thousand years ago) essentially summarizes my perception of what is going on in econometrics. Dynamic economic modelling is a comprehensive term. It covers everything except pure cross-section analysis. Hence, I have to narrow down the scope of my paper. I shall not cover duration models, event studies, count data and Markovian models. The areas covered are: dynamic panel data models, dynamic models with limited dependent variables, unit roots, cointegration, VAR’s and Bayesian approaches to all these problems. These are areas I am most familiar with. Also, the paper is not a survey of recent developments. Rather, it presents what I feel are important issues in these areas. Also, as far as possible, I shall relate the issues with those considered in the work on Political Methodology. I have a rather different attitude towards econometric methods which my own colleagues in the profession may not share. In my opinion, there is too much technique and not enough discussion of why we are doing what we are doing. I am often reminded of the admonition of the queen to Pollonius in Shakespeare’s Hamlet, “More matter, less art.”

48
Paper
A default prior distribution for logistic and other regression models
Gelman, Andrew
Jakulin, Aleks
Pittau, Maria Grazia
Su, Yu-Sung

Uploaded 08-03-2007
Keywords Bayesian inference
generalized linear model
least squares
hierarchical model
linear regression
logistic regression
multilevel model
noninformative prior distribution
Abstract We propose a new prior distribution for classical (non-hierarchical) logistic regression models, constructed by first scaling all nonbinary variables to have mean 0 and standard deviation 0.5, and then placing independent Student-$t$ prior distributions on the coefficients. As a default choice, we recommend the Cauchy distribution with center 0 and scale 2.5, which in the simplest setting is a longer-tailed version of the distribution attained by assuming one-half additional success and one-half additional failure in a logistic regression. We implement a procedure to fit generalized linear models in R with this prior distribution by incorporating an approximate EM algorithm into the usual iteratively weighted least squares. We illustrate with several examples, including a series of logistic regressions predicting voting preferences, an imputation model for a public health data set, and a hierarchical logistic regression in epidemiology. We recommend this default prior distribution for routine applied use. It has the advantage of always giving answers, even when there is complete separation in logistic regression (a common problem, even when the sample size is large and the number of predictors is small) and also automatically applying more shrinkage to higher-order interactions. This can be useful in routine data analysis as well as in automated procedures such as chained equations for missing-data imputation.

49
Paper
Formal Tests of Substantive Significance for Linear and Non-Linear Models
Esarey, Justin
Danneman, Nathan

Uploaded 07-16-2010
Keywords statistical decision theory
substantive significance
marginal effects
bayesian
Abstract We propose a critical statistic c^{*} for determining the substantive significance of an empirical result, which we define as the degree to which it justifies a particular decision (such as the decision to accept or reject a theoretical hypothesis), and provide software tools for calculating c^{*} for a wide variety of models. Our procedure, which is built on ideas from Bayesian statistical decision theory, helps researchers improve the objectivity, transparency, and consistency of their assessments of substantive significance.

50
Paper
Time Series Cross-Sectional Analyses with Different Explanatory Variables in Each Cross-Section
Girosi, Federico
King, Gary

Uploaded 07-11-2001
Keywords Bayesian hierarchical model
time series
cross-section
Abstract The current animosity between quantitative cross-national comparativists and area studies scholars originated in the expanding geographic scope of data collection in the 1960s. As quantitative scholars sought to include more countries in their regressions, the measures they were able to find for all observations became less comparable, and those which were available (or appropriate) for fewer than the full set were excluded. Area studies scholars appropriately complain about the violence these procedures do to the political reality they find from their in depth analyses of individual countries, but as quantitative comparativists continue to seek systematic comparisons, the conflict continues. We attempt to eliminate a small piece of the basis of this conflict by developing models that enable comparativists to include different explanatory variables, or the same variables with different meanings, in the time-series regression in each country. This should permit more powerful statistical analyses and encourage more context-sensitive data collection strategies. We demonstrate the advantages of this approach in practice by showing how out-of-sample forecasts of mortality rates in 25 countries, 17 age groups, and 17 causes of death in males and 20 in females from this model out-perform a standard regression approach.


< prev 1 2 next>
   
wustlArtSci