现代中西医结合杂志
現代中西醫結閤雜誌
현대중서의결합잡지
MODERN JOURNAL OF INTEGRATED TRADITIONAL CHINESE AND WESTERN MEDICINE
2014年
35期
3877-3879,3894
,共4页
钱淑雯%李海燕%杨学智%李慧%张尚尚%芦煜%王京平%张治霞%徐静
錢淑雯%李海燕%楊學智%李慧%張尚尚%蘆煜%王京平%張治霞%徐靜
전숙문%리해연%양학지%리혜%장상상%호욱%왕경평%장치하%서정
时间序列%缺失值%填补方法
時間序列%缺失值%填補方法
시간서렬%결실치%전보방법
times series%missing value%fill method
目的:模拟同一时间点数据完全缺失和部分缺失2种情况,通过填补值和实际值的对比,比较各填补方法对褪黑素(MT)时间序列的填补效果。方法同一时间点完全缺失时,比较实际值与 SPSS 5种填补方法填补结果;部分缺失时,除完全缺失的填补方法外,增加拟合时间序列模型填充。结果完全缺失时,临近点的中位数和线性插值法的填补结果和两因素析因设计资料方差分析结果更接近于实际。实际值波动幅度较小的时候,插值法拟合效果好;在实际值波动较大时,临近中位数拟合效果好。部分缺失时,拟合模型填充效果好。结论完全缺失时,如排除缺失值大幅波动,可以运用临近中位数和插值法对缺失值进行填充。在临近值波动幅度较小时,选用插值法填充值;在临近值波动幅度较大时,选用临近中位数填充值。部分缺失时,选用时间序列拟合模型填充。
目的:模擬同一時間點數據完全缺失和部分缺失2種情況,通過填補值和實際值的對比,比較各填補方法對褪黑素(MT)時間序列的填補效果。方法同一時間點完全缺失時,比較實際值與 SPSS 5種填補方法填補結果;部分缺失時,除完全缺失的填補方法外,增加擬閤時間序列模型填充。結果完全缺失時,臨近點的中位數和線性插值法的填補結果和兩因素析因設計資料方差分析結果更接近于實際。實際值波動幅度較小的時候,插值法擬閤效果好;在實際值波動較大時,臨近中位數擬閤效果好。部分缺失時,擬閤模型填充效果好。結論完全缺失時,如排除缺失值大幅波動,可以運用臨近中位數和插值法對缺失值進行填充。在臨近值波動幅度較小時,選用插值法填充值;在臨近值波動幅度較大時,選用臨近中位數填充值。部分缺失時,選用時間序列擬閤模型填充。
목적:모의동일시간점수거완전결실화부분결실2충정황,통과전보치화실제치적대비,비교각전보방법대퇴흑소(MT)시간서렬적전보효과。방법동일시간점완전결실시,비교실제치여 SPSS 5충전보방법전보결과;부분결실시,제완전결실적전보방법외,증가의합시간서렬모형전충。결과완전결실시,림근점적중위수화선성삽치법적전보결과화량인소석인설계자료방차분석결과경접근우실제。실제치파동폭도교소적시후,삽치법의합효과호;재실제치파동교대시,림근중위수의합효과호。부분결실시,의합모형전충효과호。결론완전결실시,여배제결실치대폭파동,가이운용림근중위수화삽치법대결실치진행전충。재림근치파동폭도교소시,선용삽치법전충치;재림근치파동폭도교대시,선용림근중위수전충치。부분결실시,선용시간서렬의합모형전충。
Objective It is to simulate two cases of data completely missing and partially missing at the same time, and to compare the fill effect of melatonin (MT) time series of every method by comparing the fill value and the actual value.Meth-ods In the case of data completely missing at the same time, five SPSS fill methods were used to fill the simulated missing val-ues.Filling results were compared with the actual value.In the case of data partially missing at the same time, in addition to the completely missing filling methods, the time series model fitting was added to fill.Res ults In the case of data completely missing at the same time, median fill method of approaching points and linear interpolation results were closer to the actual val-ue.Two-factor factorial design analysis of variance results were better.When the actual value had small fluctuations, the line-ar interpolation result was closer to the actual value.When the actual value had small fluctuations, the linear interpolation re-sult was closer to the actual value.When the actual value had the larger fluctuations, the median fill method of approaching points result was closer to the actual value.In the case of data partially missing at the same time, the time series model fitting had the better result.Co nclusion In the case of data completely missing, excluding the very large increase and decrease, me-dian fill method of approaching points and linear interpolation can be used to fill missing values.When the actual value has small fluctuations, choose linear interpolation.When the actual value has the larger fluctuations, choose median fill method of approaching points.In the case of data partially missing, choose time series model to fill.