论文部分内容阅读
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。
Objective To simulate two cases of data loss and partial deletion at the same time, and to compare the effect of each method of filling on the melatonin (MT) time series by comparing the filling value with the actual value. When the method is completely missing at the same time point, the actual value is compared with the SPSS five kinds of filling methods to fill the results. When the partial missing method is added, the fitting time series model is added in addition to the complete missing method. When the result is completely missing, the median of adjacent points and the result of interpolation by linear interpolation and the ANOVA of two factorial factorial design data are closer to reality. Interpolation method fits well when the actual value fluctuates less; when the actual value fluctuates greatly, the near median fitting effect is good. Part of the missing, the fitting model filling effect is good. When the conclusion is completely missing, if the missing value is greatly fluctuated, the missing value can be filled by using the median and the interpolation method. In the vicinity of the smaller fluctuations, the choice of interpolation method to fill the value; in the vicinity of the larger fluctuations, the use of the median near the fill value. Part of the missing, the use of time series fitting model to fill.