Statistical interactions in studies of physician utilization. Promise and pitfalls.
Statistical interactions in studies of physician utilization. Promise and pitfalls.
复制标题
医生利用研究中的统计交互作用。
DOI：
10.1097/00005650-198804000-00005
复制
发表时间：
1988
期刊：
影响因子：
3
通讯作者：
Harrison,KA
中科院分区：
文献类型：
--
作者：
Ronis,DL;Harrison,KA
有人建议，使用交互式统计模型将大大增加医生利用研究的方差占的比例。本文的目的是评估和描述的好处和陷阱，使用交互式统计模型的医生利用。本文介绍了蒙特卡罗模拟数据和真实的世界数据，以确定有多少医生利用的方差可以占互动回归模型。结果表明，增加交互作用项是不可能产生大的方差增加占。当预测变量中存在大量测量误差时，交互式模型的有用性特别低。交互模型的其他优点和缺点进行了讨论，包括1）提高理解，2）通货膨胀的阿尔法，3）敏感性的转换和规模的测量，和4）混杂的相互作用效应与非线性效应。
It has been suggested that use of interactive statistical models would greatly increase the proportion of variance accounted for by studies of physician utilization. The purpose of this paper is to evaluate and describe the benefits and pitfalls of using interactive statistical models of physician utilization. The paper presents Monte Carlo simulation data and real world data to determine how much more of the variance in physician utilization can be accounted for by interactive regression models. Results indicate that adding interaction terms is unlikely to produce large increases in variance accounted for. The usefulness of interactive models is particularly low when there is substantial measurement error in the predictor variables. Other advantages and disadvantages of interactive models are discussed, including 1) improved understanding, 2) inflation of alpha, 3) sensitivity to transformations and scale of measurement, and 4) confounding of interaction effects with nonlinear effects.