Binary Regression Models for Contaminated Data
J. B. Copas · Journal of the Royal Statistical Society Series B (Statistical Methodology) · 1988
SUMMARY The nature of outliers in the context of binary regression data is discussed. Resistant fitting procedures produce estimated regression coefficients which are numerically larger than maximum likelihood, implying the need for a shrinkage-type correction. A simple model which allows for a small number of the binary responses being misrecorded is proposed, and associated techniques for robust estimation and diagnostics developed.
Cited by 262 works. doi:10.1111/j.2517-6161.1988.tb01723.x
Authors
References
- Maximum Likelihood from Incomplete Data Via the EM Algorithm
- Generalized Linear Models
- Theoretical Statistics
- Robust Statistics: The Approach Based on Influence Functions
- Generalized Linear Models
- Estimation of Exponential Survival Probabilities with Concomitant Information
- Residuals and Influence in Regression
- A REVIEW OF GOODNESS OF FIT STATISTICS FOR USE IN THE DEVELOPMENT OF LOGISTIC REGRESSION MODELS1
- Logistic Regression Diagnostics
- Robust Statistics: The Approach Based on Influence Functions.