Unifying Political Methodology

by King, Gary

• ISBN13: 9780472085545
• ISBN10: 0472085549
• Edition: Reprint
• Format: Paperback
• Copyright: 1998-07-01
• Publisher: Univ of Michigan Pr

Summary

One of the hallmarks of the development of political science as a discipline has been the creation of new methodologies by scholars within the discipline--methodologies that are well-suited to the analysis of political data. Gary King has been a leader in the development of these new approaches to the analysis of political data. In his book,Unifying Political Methodology, King shows how the likelihood theory of inference offers a unified approach to statistical modeling for political research and thus enables us to better analyze the enormous amount of data political scientists have collected over the years. Newly reissued, this book is a landmark in the development of political methodology and continues to challenge scholars and spark controversy. "Gary King'sUnifying Political Methodologyis at once an introduction to the likelihood theory of statistical inference and an evangelist's call for us to change our ways of doing political methodology. One need not accept the altar call to benefit enormously from the book, but the intellectual debate over the call for reformation is likely to be the enduring contribution of the work." --Charles Franklin,American Political Science Review "King's book is one of the only existing books which deal with political methodology in a clear and consistent framework. The material in it is now and will continue to be essential reading for all serious students and researchers in political methodology." --R. Michael Alvarez, California Institute of Tech-nology Gary King is Professor of Government, Harvard University. One of the leading thinkers in political methodology, he is the author ofA Solution to the Ecological Inference Problem: Reconstructing Individual Behavior from Aggregate Dataand other books and articles.

Table of Contents

Preface	xi
I Theory	3	(94)
1 Introduction	3	(11)
1.1 Toward a new political methodology	3	(3)
1.2 A language of inference	6	(7)
1.3 Concluding remarks	13	(1)
2 Conceptualizing uncertainty and inference	14	(24)
2.1 Probability as a model of uncertainty	14	(2)
2.2 Inverse probability as a failed model of inference	16	(5)
2.3 Likelihood as a model of inference	21	(7)
2.4 The Bayesian model of inference	28	(2)
2.5 Restrictive versus unrestrictive statistical models	30	(5)
2.6 Specification tests	35	(1)
2.7 Concluding remarks	36	(2)
3 The probability model of uncertainty	38	(21)
3.1 The probability model	38	(3)
3.2 Univariate probability distributions	41	(16)
Bernoulli distribution	42	(1)
Binomial distribution	43	(2)
Extended beta-binomial distribution	45	(3)
Poisson distribution	48	(3)
Negative binomial distribution	51	(2)
Normal distribution	53	(1)
Log-Normal distribution	54	(2)
Where derivation from first principles is difficult or indeterminate	56	(1)
3.3 Multivariate probability distributions	57	(1)
3.4 Concluding remarks	58	(1)
4 The likelihood model of inference	59	(38)
4.1 Likelihood and summary estimation	59	(7)
4.2 Likelihood and point estimation	66	(7)
Analytical methods	67	(5)
Numerical methods	72	(1)
4.3 An alternative representation of the likelihood function	73	(1)
4.4 Properties of maximum likelihood estimators	74	(7)
Regularity conditions	74	(1)
Finite sample properties	75	(2)
Asymptotic properties	77	(4)
4.5 Problems to avoid with maximum likelihood estimators	81	(2)
4.6 Precision of maximum likelihood estimators	83	(9)
Likelihood ratio test	84	(3)
Direct measures of precision	87	(3)
Wald's test	90	(2)
4.7 Likelihood and interval estimation	92	(1)
4.8 Concluding remarks	93	(4)
II Methods	97	(158)
5 Discrete regression models	97	(36)
5.1 Binary variables	98	(4)
5.2 Interpreting functional forms	102	(8)
Graphical methods	104	(2)
Fitted values	106	(1)
First differences	107	(1)
Derivative methods	108	(2)
5.3 Alternative justifications for binary variable models	110	(5)
Threshold models	110	(3)
Utility maximization models	113	(1)
First principles of first principles	114	(1)
5.4 Ordered categorical variables	115	(2)
5.5 Grouped uncorrelated binary variables	117	(2)
5.6 Grouped correlated binary variables	119	(2)
5.7 Counts of uncorrelated events	121	(3)
5.8 Counts of uncorrelated events with unequal observation intervals	124	(2)
5.9 Counts of correlated events	126	(5)
5.10 Concluding remarks	131	(2)
6 Models for tabular data	133	(29)
6.1 Notation	135	(1)
6.2 The log-odds (logit) model	136	(7)
6.3 A specification test	143	(3)
6.4 The log-proportion model	146	(3)
6.5 The linear-proportion model	149	(3)
6.6 The log-frequency (log-linear) model	152	(6)
6.7 The log-odds model as a special case of the log-frequency model	158	(2)
6.8 Concluding remarks	160	(2)
7 Time series models	162	(27)
7.1 Stochastic explanatory variables	165	(2)
7.2 The influence of history	167	(9)
Exogenous variables	169	(1)
Past expectations	170	(3)
Past realizations	173	(1)
Past shocks	173	(3)
Combinations	176	(1)
7.3 Theoretical ambiguities	176	(5)
7.4 The first differences error correction model	181	(4)
7.5 The "standard" time series regression model	185	(2)
7.6 Concluding remarks	187	(2)
8 Introduction to multiple equation models	189	(19)
8.1 Identification	191	(6)
Example 1: Flat likelihoods	192	(1)
Example 2: Nonunique reparameterization	193	(1)
Example 3: Deterministic relationships among the ML estimators	194	(2)
What to do	196	(1)
8.2 Reciprocal causation	197	(4)
A linear model	198	(3)
8.3 Poisson regression models with unobserved dependent variables	201	(6)
8.4 Concluding remarks	207	(1)
9 Models with nonrandom selection	208	(23)
9.1 Censoring	208	(2)
9.2 Stochastic censoring	210	(3)
9.3 Stochastic truncation	213	(3)
9.4 Truncated and variance function event count models	216	(6)
A truncated negative binomial model	218	(3)
Truncated negative binomial with variance function	221	(1)
9.5 Hurdle event count models	222	(7)
9.6 Concluding remarks	229	(2)
10 General classes of multiple equation models	231	(19)
10.1 Factor analysis	231	(4)
10.2 Analyzing covariance structures	235	(3)
10.3 A specification test	238	(1)
10.4 A general linear structural equation model	239	(9)
The general model	240	(2)
The likelihood function	242	(2)
Special cases	244	(3)
General comments	247	(1)
10.5 Concluding remarks	248	(2)
11 Conclusions	250	(5)
References	255	(14)
Index	269