A Solution to Separation in Binary Response Models

By Christopher Zorn. Political Analysis, 2005

286 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/c118f67ded5dff7edb64f37be1b91afaf858be87

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Statistics and Probability

Citations in the corpus

Estimating logit models with small samples (2021, unknown): “To avoid an unfair comparison, we exclude the ML estimates where separation occurs (Zorn, 2005).”
Maximum Likelihood and Firth Logistic Regression of the Pedestrian Route Choice (2017, background): “Also, Zorn (2005) performed an extensive review and provided separation examples to show the guaranteed finiteness of the penalized likelihood estimates.”
Earthquakes, Religion, and Transition to Self-Government in Italian Cities (2016, background): “More generally, all the observations for which the independent variable perfectly predicts the transition outcome would be dropped from the analysis (see Zorn, 2005).”
Evaluating the Nuclear Peace Hypothesis (2009, unknown): “…consequences of proliferation (Berkowitz 1985; Brito and Intriligator 1996; Bueno de Mesquita and Riker 1982; Harvey and James 1996; Intriligator and Brito 1981; Langlois 1991; Nalebuff 1988; Powell 1985, 1987, 1988, 1989a, 1989b, 1990; Schelling 1960, 1966; Wagner 1991; Zagare and Kilgour 2000).”
Bias reduction in exponential family nonlinear models (2009, unknown): “2 Outline The recent applied and methodological interest (Mehrabi & Matthews, 1995; Heinze & Schemper, 2002; Bull et al., 2002, 2007; Zorn, 2005) motivates the in-depth study of the bias-reduction method (Firth, 1993) and the derivation of general expressions for the modified score functions in the ”
Multilevel Analysis with Few Clusters: Improving Likelihood-Based Methods to Provide Unbiased Estimates and Accurate Inference (2020, background): “That is, coefficient estimates tend to be systematically and substantially larger (in absolute size) than the corresponding true values when the sample size is small; see Firth (1993); Kosmidis and Firth (2009); Zorn (2005).”
The Global Financial Crisis and currency crises in Latin America (2012, methodology): “The first uses dynamics factors calculated from the data set, excluding institutional variables, because the low variation of some discrete variables (particularly institutional variables) may cause quasi complete separation (Zorn, 2005).”
The Political Economy of Budget Deficits (2007, unknown): “This is the problem of quasi-separation that may impede maximum-likelihood estimation even in large samples (Zorn, 2005).”
The innovative performance of firms in heterogeneous environments: the interplay between external knowledge and internal absorptive capacities (2018, background): “(1) through binary response models – such as probit models – is subject to separation problems (Zorn, 2005), which arise when one or more covariates perfectly predict the outcome of interest and is intuitive when both the dependent and independent variables of interest are binary.13 The…”
Jeffreys-prior penalty, finiteness and shrinkage in binomial-response generalized linear models (2018, methodology): “The particularly strong uptake of the method in health and medical sciences and in politics stems largely from the works of Heinze & Schemper (2002) and Zorn (2005), respectively.”
Comparing Random Forest with Logistic Regression for Predicting Class-Imbalanced Civil War Onset Data (2016, methodology): “To correct for the bias arising from class-imbalanced data, a correction to the likelihood function should be utilized.”
On the Use of Cauchy Prior Distributions for Bayesian Logistic Regression (2015, methodology): “Removing certain covariates from the regression model may appear to be an easy remedy for the problem of separation, but this ad-hoc strategy has been shown to often result in the removal of covariates with strong relationships with the response (Zorn, 2005).”
How do activities conducted while commuting influence mode choice? Using revealed preference models to inform public transportation advantage and autonomous vehicle scenarios (2019, background): “We speculate that the difficulties we encountered in finding stable specifications of the nested logit model are probably in large measure a result of complete or quasicomplete separation problems arising from the unbalanced shares (Zorn, 2005).”