8.数据分析方法

在撰写数据分析方法这一部分时，注意内容的完整性和相关性。一般多包含以下几个方面：①数据分析软件（标明版本、软件公司及国家、地区）；②如果是meta-analysis，需要给出具体的数据量化整合方法，例如：以森林图的形式展示各研究及汇总后（pooled）的结果；③说明缺失值的处理方法（例如，根据已有数据计算而来、通过联系作者获得等）；④说明是否进行亚组分析（以及进行此分析的依据）、敏感性分析以及发表偏倚分析。此外，对于无法进行量化的结果，可采用质性方式进行总结，也需对此进行说明。以下将对几个案例着重进行介绍。如示例5-21所示，该综述根据干预类型进行了亚组分析。如示例5-22所示，该综述未进行亚组分析，并说明了相应的原因。如示例5-23所示，该综述对发表偏倚进行了评估，并进行了敏感性分析。

示例5-21　亚组分析

Subgroup analyses were conducted based on the type of intervention：total sleep deprivation（TSD）where participants were completely deprived of any sleep or partial sleep restriction（PSR）where participants'sleep was reduced relative to a basal amount.Additional subgroup analysis was conducted based on sex and the methods of assessment（wherever applicable）.

示例5-22　无亚组分析及其原因(https://www.daowen.com)

For this paper，we also considered the type of subgroup analysis performed in the Meekums et al.（2015）study based on age，but chose not to pursue this mainly due to the fact that only one study survived scrutiny that did not involve adults；this one study was also assessed as having high risk of bias.Subgroup analysis for the type of intervention was not performed either because there were no obvious differences between DMT practices used in the reviewed studies.Finally，although there were differences in the severity of depression at the start of the study，subgroup analysis on the level of depression was not performed because of the limited number of studies；differences were discussed in the narrative metasynthesis.

示例5-23　发表偏倚及敏感性分析

Publication bias or reporting bias was assessed using the funnel plot alongside the Begg's Test and Eggers'Test.Sensitivity analysis was examined by the Jackknife Method.In this method，which is also known as the‘one out method’，the quality and consistency of the results were evaluated by removing each study individually.