Out today in @PNASNews: "Neuroticism May Not Reflect Emotional Variability", a joint-first author effort from me and @seanchrismurphy, also including @drbrockbastian, @tomhollenstein (and others not on twitter): https://www.pnas.org/content/early/2020/04/15/1919934117. Thread to follow (1/12)
The inverse of neuroticism is "emotional stability". Almost all neuroticism scales have items assessing emotional variability (e.g. "up and down mood swings"). So it's perhaps not surprising that many studies link neuroticism with emotional variability (2/12)
BUT these findings may be because of a measurement issue. There can be dependency between variability and mean levels, particularly when using bounded scales. For e.g. on a 0-100 scale, if my mean is 10, I can't show as much variability as somebody with a mean of 50 (3/12)
This issue is compounded for then neuroticism-emotional variability relationship, because 1) variability is calculated using emotion assessed in everyday life, where negative emotion is often at floor (at least, pre-COVID!) - this figure shows this skew in our datasets (5/12)
2) Low neuroticism is strongly linked with low mean levels of negative emotion, meaning that those low in neuroticism may be systematically precluded from demonstrating variability (6/12)
We tested this by meta-analyzing the neuroticism-negative emotional variability association in 11 experience-sampling datasets. We found the expected association. We tested whether it held across 2 analyses aiming to disentangle variability from mean levels (7/12)
1) To remove many of the scores at scale floor, we calculated variability in daily max negative emotion (in the 8 studies that had multiple measurements per day). Negative emotional variability was still associated with neuroticism, but the association was attenuated (8/12)
2) We used the relative variability index developed by @merijn_mestdagh, which calculates variability as a proportion of the max variability possible given the mean. When using this index, neuroticism was no longer associated with negative emotional variability (9/12)
That is, once we corrected dependency between variability and the mean, neuroticism wasn't characterized by emotional variability. Thus, the mean and variability reflect the same info, and given the mean is more parsimonious, we think there's no need to invoke variability (10/12)
This suggests we should understand what's going on with the mean-variability relationship before interpreting variability as meaningful. We also need to wrestle with emotion measurement: does the scale floor truly mean "no negative emotion" to Ps? (11/12)
If you want to know more, here's a link to a preprint of the paper: http://elisekalokerinos.com/wp-content/uploads/2020/04/Kalokerinos-Murphy-et-al.-2020-PNAS-Neuroticism-variability.pdf, and here's a link to our data and code: https://osf.io/gvfdx/ (12/12)