Neuroticism may reflect emotional variability when correcting for the confound with the mean

Kalokerinos et al. (1) demonstrate that emotional variability, as assessed via the SD, is not significantly associated with neuroticism once the confound with mean negative emotion is controlled for, r = 0.05. To control for this confound, the authors use the relative SD (RSD), which mathematically corrects for the nonlinear dependency between the SD and the mean (2). The authors prefer the RSD over including the SD and the mean of negative emotions in a model, given the nonlinearity of their association. While we agree that the RSD is an important addition to the field, we argue that including the mean of emotions is still necessary when examining the association between neuroticism and emotion variability, to 1) sufficiently control for the confound with the mean and 2) compare the individual contribution of RSD and mean negative emotion when predicting neuroticism.

The authors (1) state that the RSD controls for the confound with the mean. However, they do not report an empirical test of this claim. Using the 11 datasets the authors provide, we find that the RSD is not significantly associated with mean negative emotion (Fig. 1A) and positive emotion (Fig. 1B). However, variability in the effect sizes is very high, and individual effect sizes for negative emotion range considerably. A metaregression of the effect sizes of the association between neuroticism and RSD reveals that the individual correlations between RSD and mean negative emotion completely explain the heterogeneity of the effect sizes of the neuroticism and RSD association, b = 0.33, SE = 0.12, 95% CI [0.10, 0.56], R² = 100%.

Fig. 1.

Forest plots of the relationship between relative negative emotional variability and (A) mean negative emotion, (B) mean positive emotion, and (C) neuroticism when controlling for mean negative and mean positive emotion. (D) Relationship between mean negative emotion and neuroticism, controlling for relative variability in negative emotion and mean positive emotion. For each dataset, we provide a standardized coefficient bounded by a 95% CI. The area of each square is proportional to the weight of the study in the metaanalysis. The results of the random effects metaanalysis are depicted at the bottom of each plot (RE model), with the width of the rhombus representing the 95% CI. The dotted line represents no effect. DD, daily diary; ESM, experience sampling method.

Thus, we conclude that including mean emotion is still important when examining emotional variability, given the remaining confound with the mean. Moreover, the association between neuroticism and mean negative emotion is not controlled for the confound with the RSD. Thus, to quantify the individual contribution of RSD and mean negative emotion, we reanalyzed the data (3), transformed the right-skewed variables, and performed linear regressions on each dataset where we predicted neuroticism by RSD and mean negative and mean positive emotion (analysis script, data, and results: https://osf.io/4c23g/). This yielded a significant mean effect size of β = 0.10, 95% [0.05, 0.15] for RSD (Fig. 1C) and β = 0.27, 95% [0.21, 0.33] (Fig. 1D), amounting to β = 0.14 and β = 0.33, respectively, when Spearman−Brown correcting the individual effect sizes to control for the unreliability of the measures. Importantly, adding the RSD to the model with mean emotion improved the model fit (Akaike information criterion = 3,049.9 compared to 3,062.1) and increased the explained total variance from $R_t^{2\left(f\right)}$ = 15.4 to 17.5% (4). Although only small in size according to general guidelines (5, 6), it can be viewed as a substantial increase given the broad construct of the Big Five personality traits (7) that often lead to small correlations with narrower constructs (e.g., refs. 8 and 9). To conclude, our reanalysis shows that, although mean negative emotion is more strongly associated with neuroticism, emotional variability provides unique information about neuroticism above and beyond mean emotion.