摘要
目的 阐明回归树现有算法应用于病案首页资料分析时的适用条件 ,并提出其替代算法。方法 从回归树现有算法的原理上阐明了当应变量分布非正态时可能出现的问题 ,并采用病案资料加以验证。结果 当应变量非正态分布时 ,直接应用现有回归树算法不能得出正确的分析结论 ;而当存在极端值时这一问题更为严重。结论 虽然树模型是非参数分析方法 ,但回归树仍有其适用条件 ,在使用时需要对这些条件加以检查 ,必要时应采取相应的处理措施。
Objective To explore the condition of application for regression tree in analysis of inpatient data,meanwhile to discuss its surrogate arithmetic. Methods Based on its principle,underlying problem for regression tree are explicated,inpatient cost data are applied to confirm the problem.Results Default arithmetic of regression tree can't be used directly under the condition of non-normal distribution,existence of outlier would make the problem more severity. Conclusion Regression tree also has its application condition,these condition should be checked before analysis,correspond step should be applied when necessary.
出处
《中国卫生统计》
CSCD
北大核心
2003年第6期338-340,共3页
Chinese Journal of Health Statistics
基金
复旦大学青年教师科研启动基金资助项目
