Relaxing the rule of ten events per variable in logistic and Cox regression.

By E. Vittinghoff, Charles E. McCulloch. American Journal of Epidemiology, 2007

4,006 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/2acfe2681d042e872827e7caab22018079f51153

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

Revisiting the past to foretell the future: summer temperature and habitat area predict pika extirpations in California (2015, methodology): “To avoid over-fitting we restricted all models to a maximum of five events per variable (Vittinghoff & McCulloch, 2007).”
Reporting Recommendations for Tumor Marker Prognostic Studies (REMARK): Explanation and Elaboration (2012, background): “In this situation sample size need not be as large as the 10-25 events per variable rule would recommend [90] and other sample size calculation methods that appropriately account for correlation of the marker with the other variables are available [82,91].”
The number of subjects per variable required in linear regression analyses. (2015, methodology): “In a more recent study, Vittinghoff and McCulloch [10] found that this requirement of 10 EPV could be relaxed in the context of confounder adjustment, to requiring ﬁve to nine EPV.”
Developing prediction models for clinical use using logistic regression: an overview. (2019, methodology): “However, this method has largely been replaced as models with higher and lower EVP are all susceptible to bias (17).”
Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modelling Studies: The CHARMS Checklist (2014, methodology): “The number of events-pervariable (EPV) is commonly used to calculate the sample size, where attaining a sample size with an EPV of ten or more is frequently recommended to avoid overfitting [97–101].”
Minimum sample size for developing a multivariable prediction model: PART II ‐ binary and time‐to‐event outcomes (2018, methodology): “When Nagelkerke's R2, ln Lnull, and n are available, the R(2)CS_app can be calculated by rearranging Equation (14) to give”
Prognosis and prognostic research: what, why, and how? (2009, background): “although a recent study showed that this number could be lower in certain circumstances.”
A modern maximum-likelihood theory for high-dimensional logistic regression (2018, background): “A line of simulation-based results exists to guide practitioners about the sample size required to avoid finite-sample problems (36, 37).”
Comparison of Propensity Score Methods and Covariate Adjustment: Evaluation in 4 Cardiovascular Studies. (2016, background): “However, more recent opinions favor relaxing this rule of thumb([3]).”