經(jīng)驗取樣法的數(shù)據(jù)分析:方法及應(yīng)用

摘要： 經(jīng)驗取樣法是通過對調(diào)查對象多次重復(fù)測量進(jìn)行數(shù)據(jù)收集的研究方法 ,近年來受到研究者廣泛關(guān)注。本文主要基于《應(yīng)用心理學(xué)期刊》(Journal of Applied Psychology)2010～2017年刊發(fā)的34篇文獻(xiàn),總結(jié)并述評經(jīng)驗取樣數(shù)據(jù)分析中的四個關(guān)鍵成分及具體操作:數(shù)據(jù)的結(jié)構(gòu)設(shè)置、數(shù)據(jù)清理、所用測量工具的信效度檢驗、假設(shè)檢驗方法。具體包括,數(shù)據(jù)由于重復(fù)抽樣而形成不同的嵌套結(jié)構(gòu);數(shù)據(jù)清理涉及異常值、缺失值的識別與處理;信效度計算方法區(qū)別于一般研究;假設(shè)檢驗時,依據(jù)研究問題("變量之間的關(guān)系"與"變量隨時間的變化")選擇相應(yīng)的模型構(gòu)建和估計方式。此外,收集國內(nèi)期刊刊發(fā)的12篇實證文獻(xiàn),將其與國際期刊中經(jīng)驗取樣數(shù)據(jù)的分析步驟及操作方法進(jìn)行比較。最后,對未來研究如何豐富及完善數(shù)據(jù)分析過程作了展望。
Experience sampling is an effective method of collecting longitudinal data, which draws much attention in recent years. We reviewed the data analysis process from five key steps, including data structure identifying, data cleaning, reliability and validity testing, and hypothesis testing. Each step was illustrated with examples from 34 empirical studies in Journal of Applied Psychology in recent 8 years. Specifically, data structures differ based on the sampling unit. Data cleaning involving dealing with abnormal data and missing data. Methods for testing reliability and validity differs from common studies. The selection of analytical approaches to hypothesis testing is contingent on the types of research questions. Besides, we further collected 12 empirical studies published in Chinese journals and compared the data analysis between studies of Journal of Applied Psychology and Chinese journals. Finally, the future studies of experience sampling data analysis were discussed as well.