西南交通大学学报
西南交通大學學報
서남교통대학학보
JOURNAL OF SOUTHWEST JIAOTONG UNIVERSITY
2014年
5期
824-830
,共7页
任挺%陈林%杨飞%晏启鹏
任挺%陳林%楊飛%晏啟鵬
임정%진림%양비%안계붕
渐进协方差矩阵%正交设计%随机因子设计法%意愿调查%拟合优度
漸進協方差矩陣%正交設計%隨機因子設計法%意願調查%擬閤優度
점진협방차구진%정교설계%수궤인자설계법%의원조사%의합우도
asymptotic variance-covariance matrix%orthogonal design%random factorial design%stated preference questionnaires%goodness of fit
为克服现有意愿调查(stated preference,SP)实验设计方法标定结果精度低的缺点,以Logit模型的渐进协方差矩阵行列式最小为目标,对选择枝属性水平组合进行迭代计算,得到相应的选择枝属性水平组合方案,设计了一种新的出行意愿调查实验设计方法.以某城市中心城区到卫星城新增地铁交通方式为例,通过分析拟合优度、参数估计标准差,对比了新方法与正交设计法的模型标定结果精度.算例结果表明,与正交设计法相比,采用本文方法得到的参数估计标准差最小,拟合优度达到0.15,提高了0.05,提高了Logit模型参数标定精度,且具有选择枝属性组合数量弹性选择的优点.
為剋服現有意願調查(stated preference,SP)實驗設計方法標定結果精度低的缺點,以Logit模型的漸進協方差矩陣行列式最小為目標,對選擇枝屬性水平組閤進行迭代計算,得到相應的選擇枝屬性水平組閤方案,設計瞭一種新的齣行意願調查實驗設計方法.以某城市中心城區到衛星城新增地鐵交通方式為例,通過分析擬閤優度、參數估計標準差,對比瞭新方法與正交設計法的模型標定結果精度.算例結果錶明,與正交設計法相比,採用本文方法得到的參數估計標準差最小,擬閤優度達到0.15,提高瞭0.05,提高瞭Logit模型參數標定精度,且具有選擇枝屬性組閤數量彈性選擇的優點.
위극복현유의원조사(stated preference,SP)실험설계방법표정결과정도저적결점,이Logit모형적점진협방차구진행렬식최소위목표,대선택지속성수평조합진행질대계산,득도상응적선택지속성수평조합방안,설계료일충신적출행의원조사실험설계방법.이모성시중심성구도위성성신증지철교통방식위례,통과분석의합우도、삼수고계표준차,대비료신방법여정교설계법적모형표정결과정도.산례결과표명,여정교설계법상비,채용본문방법득도적삼수고계표준차최소,의합우도체도0.15,제고료0.05,제고료Logit모형삼수표정정도,차구유선택지속성조합수량탄성선택적우점.
In order to overcome the low precision of the model estimation results in the existing stated preference (SP)experimental designs,a new experimental design method for SP survey was proposed. This method aims to minimize the determinant of asymptotic variance-covariance (AVC )matrix to compute iteratively the alternative attribute level combinations and then obtain the corresponding alternative attribute level combinations. A new metro connecting a city and its satellite cities was taken as the research case. Comparisons between the new design and the orthogonal design are made about the precision of model results by the goodness of fit and standard deviation of parameter estimations. The results indicate that based on the new design,the standard deviation of parameter estimations is less and the goodness of fit is 0 . 145 ,which is improved about 0 . 05 compared to the orthogonal design. In addition,the new design improves the precision of Logit model parameter estimation and has the advantage of the flexibility in choosing the number of combinations.