Kowalski C J, Schneiderman E D, Willis S M
Department of Biologic and Materials Sciences, Dental School, University of Michigan, Ann Arbor 48109, USA.
Comput Biol Med. 1994 Nov;24(6):485-92. doi: 10.1016/0010-4825(94)90046-9.
A menu-driven PC program implementing Blomqvist's [J. Am. Stat. Assn. 72, 746-749, 1977] method for assessing the relationship between rate of change and initial value is described, illustrated and made available. It is shown that the naive approach to this problem--computing the correlation between the initial value and either the amount or rate of change--results in a negatively biased estimator. The extent of this bias can be dramatic and may lead investigators to conclude that a negative correlation is present when none exists; or that there is no correlation when in fact the correlation is positive. Blomqvist's (maximum likelihood) estimator avoids this bias, and is obtained by a transformation of the naive estimator.
描述、说明并提供了一个菜单驱动的个人计算机程序,该程序实现了布洛姆奎斯特[《美国统计协会杂志》72, 746 - 749, 1977]评估变化率与初始值之间关系的方法。结果表明,处理该问题的简单方法——计算初始值与变化量或变化率之间的相关性——会产生负偏差估计量。这种偏差的程度可能很大,可能导致研究人员得出不存在负相关时存在负相关的结论;或者在实际上存在正相关时得出不存在相关性的结论。布洛姆奎斯特(最大似然)估计量避免了这种偏差,并且是通过对简单估计量进行变换得到的。