應用一日重建法和分類與迴歸樹探討主觀幸福感

Abstract

本研究旨在運用新穎的主觀幸福感（subjective well-being）調查方法一日重建法（Day Reconstruction Method, DRM）及非線性的分類與迴歸樹（Classification and Regression Tree, CART）統計方法，探究主觀幸福感的情緒面。針對100名大學生為對象，用DRM蒐集到1657個以事件為單位的情緒資料，再運用CART統計方法，分析事件的情境與情緒間的關係。 分析資料時，DRM能得知受試者在不同情境變項下的情緒總和分數；以CART分析則能進一步提供情境變項條件組合與情緒的關係；將事件的情緒分數彙總，則可以得知受試者個人或群體的主觀幸福感分數，包含以持續時間加權的總和情緒與U指數；個人情緒可以與主觀幸福感的認知面──生活滿意度進行比較。 研究結果顯示，從情境分類求情緒平均分數而言，大學生在從事約會/親密關係、放鬆/玩遊戲、飲食/聚餐、看電視/電影/聽音樂、逛街/購物等感情交友、吃喝玩樂方面的事情有最正向情緒；學校課業、社團活動、工作/打工等事情則有最負向的情緒。以CART分析則發現，在所有事件中，大學生的情緒與學校功課、放鬆/玩遊戲、會議/團體討論、與好朋友互動、交通/移動等情境有重大關係；此外CART將樣本分割成16個終端節點，各有不同的條件可預測情緒。彙整事件情緒資料到個人層次則發現主觀幸福感的情緒面與認知面明顯不相同，呈低度相關。 本研究還發現CART分析DRM資料時的優點，包括能同時分析不同資料型態且數量眾多的情境變項、找到特定的情境變項組合、排序情境變項的重要性等。 最後，這些發現在應用方面提供一些增進大學生主觀幸福感的建議。在研究方面建議可善加利用DRM的彈性拓展未來研究方向，包含改變情境變項、受試者族群、背景變項、情緒詞，並可搭配有時間資訊的記錄工具瞭解人們生活情境與情緒的關係。此外，亦可蒐集代表性良好的樣本，以CART建立預測幸福感的模式。

The purpose of this study was to investigate the affect part of subjective well-being (SWB) through a new kindsurvey method called Day Reconstruction Method (DRM) with non-liner statistical analysis method Classification and Regression Tree (CART). In this study, 1657 episodes of affect data was collected from 100 undergraduate students using DRM, these data then analyzed by CART in order to further understand the relation between situation and affect of episodes. When analyzing DRM, one could get net affect scores from different situation variables; CART analysis provided further information of relationship of situation conditions combination and affect; if summarize episodic data, one could get individual subjective well-being scores, including duration-weighted net affect and U-index; individual affect then could be compared with cognitive part of SWB, that is, life satisfaction. Mean scores of affect from different situation showed that undergraduate students had the most positive affect when the episodes involved with friendship and playing, such as dating/intimacy relationship, relax/playing games, eating, watching TV/movies/listening to music, shopping; and had the most negative affect when the episodes involved with schoolwork, school clubs, working/part time jobs. CART analysis showed that for all episodic affect, schoolwork, relax/playing games, meeting/group discussion, interaction with good friends, traffic/moving have the most important relationships. CART analysis also divided samples into 16 terminal nodes, these nodes have different conditions, which can be used to predict affect score. When summarized episodic data to individual level, that is, the affect part of SWB, one could find it differs from the cognitive part of SWB. These two parts only had low correlation. This study also found strength of CART for DRM data analysis, including the ability of analyzing large number and different type of variables, being able to find the specific conditions combination of situation variables, sorting situation variables from importance. Finally, these findings suggested some ways to improve undergraduate students' subjective well-being. It's suggested that future research making use of the flexibility of DRM for different kinds of application, including change the situation variables, subject group, background variables, and affect terms, and use other recording tools with time information to gain understanding of relationship of situation and affect. On the other hand, one can establish well-being predicting model by CART with collecting samples with good representativeness.