## Abstract

A new stochastic framework is proposed for evaluating the individual and collective impact of cancer risk factors, and is applied to data on the incidence of melanoma. It is demonstrated that the standardized incidence ratio of second primary melanoma can be used to estimate the total coefficient of variation in risk in the population, subject to some simplifying assumptions. The coefficient of variation estimated in this manner thus can be used as a benchmark against which to judge the contributions to this total variance of individual risk factors. A nonparametric estimator of the coefficient of variation attributable to a single risk factor on the basis of data from a case-control study is derived, and its statistical properties are examined using simulations. It is shown that the categorization of a continuous risk factor can attenuate the estimate substantially, and that estimation of the joint contribution of several risk factors will usually require statistical modeling. Applying the methods to the epidemiology of melanoma, the results indicate that the known risk factors for melanoma explain only a relatively small fraction of the population variation in risk, in contrast to conventional views on this topic.

Original language | English (US) |
---|---|

Pages (from-to) | 415-426 |

Number of pages | 12 |

Journal | Journal of the American Statistical Association |

Volume | 93 |

Issue number | 442 |

DOIs | |

State | Published - Jun 1 1998 |

Externally published | Yes |

## All Science Journal Classification (ASJC) codes

- Statistics and Probability
- Statistics, Probability and Uncertainty

## Keywords

- Cancer incidence
- Coefficient of variation
- Empirical Bayes estimation
- Relative risk
- Standardized incidence ratio