靶场光学测量数据异方差性检验及修正  

Test and Adjust Heteroscedasticity of Optical Measure Data in Proof Range

作　　者：杨荣芳 周惠 

机构地区：[1]解放军92941部队

出　　处：《指挥控制与仿真》2008年第4期114-116,120,共4页Command Control & Simulation

摘　　要：为了避免靶场光学测量数据异方差性导致的普通最小二乘估计非有效、显著性检验失去意义和模型的预测失效问题,采用了图形分析、Goldfeld-Quandt和Breusch Pagan Godfrey方法检验光学测量数据异方差性,并针对光学测量数据的异方差性提出分段加权最小二乘修正的方法。通过理论分析,对某设备方位角测量数据进行实验验证,取得了残差平方数据、G-Q检验统计数据、BPG检验统计数据和分段加权最小二乘BPG统计数据。结果表明应用图形分析法对光学测量数据进行异方差性检验最直观和简捷,适合存在明显异方差性的检验,G-Q检验法不适用光学测量数据的异方差性检验,BPG检验理论完整且适合光学测量数据的异方差性检验,分段加权最小二乘方法有效合理,消除了异方差性对回归模型的影响。Figure analysis,Goldfeld-Quandt and Breusch Pagan Godfrey method to test the heteroscedasticity of optical measure data are applied,so invalidations of ordinary least square estimate and model prediction are avoided because of optical measure data heteroscedasticity. Method of weighted least square estimate to adjust the heteroscedasticity of optical measure data is presented. Theories analysis and the azimuth angle data test verification are carried on. Square of residual value, G-Q method test data, BPG method test data and BPG method test data after weighted least square estimate adjusting are obtained. The results show that Figure analysis is direct and simple adapting to visible heteroscedasticity of optical measure data, G-Q method can＇t test heteroscedasticity of optical measure data,BPG method is the best adaptive to test heteroscedasticity of optical measure data. Method of weighted least square estimate to adjust the heteroscedasticity of optical measure data is scientific and reasonable.The effects of heteroscedasticity to regressive model are avoided.

关 键 词：异方差性 加权最小二乘 G—Q检验 BPG检验 光学测量数据 

分 类 号：O241.5[理学—计算数学] O212.1[理学—数学]