Catholic University of Korea, Seoul, Korea.
Curr Opin Support Palliat Care. 2013 Mar;7(1):45-53. doi: 10.1097/SPC.0b013e32835bf28b.
Within a broader perspective on the next challenges in oncologic symptom cluster research, the objectives of this review are to examine the statistical methods that have been used to quantify and/or model the dynamic nature of symptom clustering, the methodological issues associated with those methods, and the statistical modeling techniques for the underlying mechanisms of symptom clustering.
Correlation, factor analysis, principal component analysis, and cluster analysis are analytical methods to identify symptom clusters and/or to examine the influence of symptom clusters on patient outcomes. More recent techniques include latent variable methods, such as latent profile analysis, to examine the phenotypes of symptom cluster experience and growth modeling to examine the longitudinal nature of symptom cluster experience. Future endeavors include an investigation of the underlying mechanisms of symptom clustering using longitudinal data analysis. The methodological issues include the domain of the symptoms, measurement errors, stability of the solution within the data, measurement timing, and sample size.
Each method has unique strengths and weaknesses, and the method choice should be driven by the aims and research questions of a given study.
在更广泛的肿瘤症状群研究的未来挑战视角下，本综述的目的是检验已用于量化和/或模拟症状群动态特性的统计方法、与这些方法相关的方法学问题，以及用于症状群根本机制的统计建模技术。
相关性、因子分析、主成分分析和聚类分析是用于识别症状群和/或检验症状群对患者结局影响的分析方法。更近期的技术包括潜在变量方法，如潜在剖面分析，以检验症状群体验的表型，以及生长建模，以检验症状群体验的纵向特性。未来的研究包括使用纵向数据分析来研究症状群的根本机制。方法学问题包括症状的领域、测量误差、数据内解的稳定性、测量时间和样本量。
每种方法都有其独特的优势和弱点，方法的选择应取决于特定研究的目的和研究问题。
