摘要
目的介绍用于检验定性时间序列周期性的谱封分析法,通过模拟研究评价其检验效果,并用实例数据进行验证。方法模拟不同序列长度、周期大小、状态空间维度的定性时间序列并进行谱封分析检验其周期性,通过阳性符合率、阴性符合率反映检验效果。结果研究发现序列长度越长时检验效果相对来说越好,状态数大于等于周期大小且序列长度大于状态数50倍时可以保证较高的符合率,此外序列长度一定时,状态数少、周期性小时检验效果好。结论谱封分析法在定性时间序列周期性检验应用中取得效果较好,在应用时要充分考虑序列长度和状态数、周期大小的关系以及可能存在的隐含周期大小。
Objective Making an introduction to spectral en- velope method,a frequency based technique for analyzing categorical time series, and evaluating the efficiency of this method for periodic detection by simulation study. Methods To simulate categorical time series with dif- ferent kinds of length ,period and state space and assess the efficiency by u- sing three assessment criteria, such as positive rate （ Rl ）, type I negative rate（ R2 ） and type II negative rate （R3 ）. Results The efficiency be- comes better as the length of series increasing and if the category number e- qual to period value and the length is 50 times larger than the category number the test is good enough, for small category number and period value under the same sequence length is the same. Conclusion In general, the spectral envelope method is an acceptable method to detect the periodicity of categorical time series and after making an adequate consideration of the characteristics of the series, especially the relationship between category number or period value and length, the efficiency can be controlled at a high level.
出处
《中国卫生统计》
CSCD
北大核心
2013年第3期330-333,338,共5页
Chinese Journal of Health Statistics
关键词
谱封分析法
定性时间序列
周期性检验
模拟
Spectral envelope method
Categorical timeseries
Periodic detection
Simulation study
