再探情緒多樣性指標之爭議

Abstract

Quoidbach等人（2014）採用資訊理論與生物多樣性研究常用的熵（Shannon’s entropy）來測量個體的情緒多樣性（emodiversity）。雖然曾有學者反對採用該指標來表徵情緒多樣性，但該指標仍在心理學研究中被廣泛使用。本文先探討以往對熵指標在情緒多樣性的辯論，並以實徵資料來重新評估用熵作為情緒多樣性的合理性。本研究更延伸出以往辯論較忽略的兩大缺陷：（一）該指標以李克式量尺（Likert-type scale）計算機率並無理論基礎；（二）在實徵研究常用的李克式量尺點數（例如5點至7點）下，該指標與情緒豐富度（richness）有極高的線性關聯。在實徵資料（N = 962）中，本研究分析在五點李克式量尺中不同題目數（5、10與20題）下正負向情緒之熵指標，也發現熵與情緒豐富度有極高的線性關聯（rs = .94 - .98）。此外，熵的變異可被平均數、相對標準差及二者交互作用之線性組合良好解釋（解釋變異量介於78%至92%）。這些結果皆顯示利用熵指標測量情緒多樣性仍有其效度上的疑慮。因此本文建議，當情緒是以五點李克式量尺測量時，研究者不應採用熵指標測量情緒多樣性，且該指標主要以其平均數跟相對標準差即可指稱之。

This article reexamines the previous debate on Quoidbach et al.’s (2014) entropy-based emodiversity index and evaluates its utility from theoretical and empirical aspects. We mainly focus on the consequences of applying Shannon’s (1948) entropy to Likert-type scale. From the previous debate, we extend and summarize two severe defects of the entropy-based emodiversity index: (a) calculating probability from Likert-type scale is mathematically unjustifiable; (b) there is extremely high collinearity between the entropy index and richness when the emotion measurement has limited range, which usually occurs in empirical studies. We also collected data (N = 962) and analyzed the entropy scores for positive and negative affect under 5-point scale with different numbers of items (5-, 10-, and 20-item). The results showed that under 5-point scale Quoidbach et al.’s index is strongly correlated with richness (rs ranged from .94 to .98) and that the variation of their index can be simply explained by the linear combination of mean, relative standard deviation, and their interaction term (R2s ranged from .78 to .92). Hence, the validity of using entropy to measure diversity in a 5-point scale is questionable and we advise researchers to prohibit the use of this index to represent emodiversity.