疾病空间分布趋势面模型的共线性偏倚及其测量与控制

目的探讨趋势面分析中共线性偏倚及其测量与控制方法.方法以疾病监测资料为基础,引用回归诊断方法识别趋势面模型的共线性偏倚,进一步用岭回归趋势面模型控制共线性偏倚.结果趋势面分析往往存在共线性偏倚,利用岭回归趋势面分析可以在一定程度上控制共线性偏倚.结论在作趋势面分析时应当考虑共线性偏倚对结果的影响,并设法予以控制.

Measure and Control of Multi-collinearity Bias in Trend-surface Models for Spatial Distribution of Disease

Objective To measure the multi-collinearity bias in trend-surface models and to control its impaction to the results of trend-surface analysis.Methods Based on the data of surveillance of disease, use the method of regression diagnostics to measure the multi-collinearity bias in the least square estimation trend-surface models ,and use the method of ridge regression to control the multi-collinearity bias in the least square estimation trend-surface models. Results there is multi-collinearity bias in the result of the least square estimation trend-surface model, use the method of ridge regression can control the multi-collinearity bias.Conclusion To the practice of trend-surface analysis in the study of geographical epidemiology, we should measure and control the multi-collinearity bias in trend-surface models.