风险估计相关数据(Relevant Data for Relative risk
estimate)
数据摘要：
Each datafile contains a matrix of values indexed by p1 = 0.02 to 0.98
by 0.02 (columns) and p2 = 0.02 to 0.98 by 0.02 (rows). Results for each of the four candidate methods are included within the single file with the index of “method”.
中文关键词：
算法,统计,风险,估计,模型,
英文关键词：
algorithms,statistic,risk,estimate,model,
数据格式：
TEXT
数据用途：
Design of Algorithms
数据详细介绍：
Relevant Data for Relative risk estimate
Data set information:
This file details the contents of the supplemental result files for the paper. Each datafile contains a matrix of values indexed by p1 = 0.02 to 0.98 by 0.02 (columns) and p2 = 0.02 to 0.98 by 0.02 (rows). Results for each of the four candidate methods are included within the single file with the index of “method”.
File naming convention:
{statistic}_n{sample size}.sas7bdat Where
statistic = {bias, coverage, MSE, rejection}, Sample size = {10, 25, 50, 100}, and
sas7bdat represent a SAS v7/8/9 permanent dataset.
Calculations:
For each (p1,p2) dyad and sample size (n1=n2), all possible sample realizations were enumerated. For example, if n1=n2=10, then sample configurations of (y1, y2) would include (0,0), (0,1), (0,2),…(1,0),
(1,1),…(10,10). The probability of each sample configuration was determined using a product binomial distribution, namely Prob=Prob[Bin(y1, p1)]*Prob[Bin(y2, p2)], where (p1,p2) are determined by row/column placement in the results matrix.
Bias:
The E[RR] was determined based on the observed estimate and the product binomial probability. Entries in “bias” files are Bias = E[RR] – p1/p2.
Coverage:
The 95% confidence interval for each of the four methods was computed as described in the paper. Values provided in the “coverage” files are the sum of I(covered)*Prob where I(covered)=1 if the 95% CI contains the true parameter (i.e., p1/p2), 0 else; and Prob is the product binomial probability.
MSE (Mean Squared Error):
MSE is determined as E[RR]^2 – Var[RR], where E[RR] is described above in
Bias and Var[RR] = sum_{i} ( RR_i – E[RR])^2 where i indexes the (n1+1)x(n2+2) potential samples given n1,n2.
Rejection:
Determination of rejection rates, based on confidence intervals, of a null
hypothesis of RR=1.0 for all (p1,p2) pairings. When p1=p2, the interpretation is a type I error. In contrast, when p1 <> p2, interpretation is of power. Source:
Rickey E. Carter
Department of Health Sciences Research Mayo Clinic
200 First Street South West Rochester MN 55905 USA
Relevant Papers:
Relative risk estimated from the ratio of two median unbiased estimates, by R. E. Carter et al., pages 657–671;
数据预览：
点此下载完整数据集
