Estimating the correlation of bivariate failure times under censoring.

摘要: The analysis of correlations within pairs survival times is interest to many research topics in medicine, such as the correlation survival-type endpoints twins, till failure paired organs, or time with a surrogate endpoint. dependence assumed monotonic and thus quantification by rank coefficients appropriate. typical censoring requires more involved methods estimation inference have been developed recent years. paper focuses on semiparametric approaches, particular normal copula-based Spearman coefficients. copula approach, often presented for mathematically inclined readership, reviewed from viewpoint an applied statistician. As alternative maximum likelihood methodology approach (NCE) we introduce iterative multiple imputation (IMI) method which only about 0.05% computing NCE, without sacrificing statistical performance. For IMI, probabilities at death are first transformed deviates. Then, those deviates that relate censored iteratively augmented, using conditional imputation, until convergence obtained scores correlation, similar Spearman's correlation. Statistical properties NCE IMI compared means Monte Carlo study three real data sets, also give impression range applications, their problems.