A Latent Variable Approach to Affect Variability in Daily Life Accurately Predicts Psychopathology, Especially Depression Symptoms in a Non-Clinical Sample.

Background: Ecological momentary assessments (EMA) have contributed to an increase in research correlating affect dynamics to mental health and wellbeing. While many metrics can be calculated to characterize affect dynamics from EMA data, researchers often opt for a 'battle royale' approach whereby only the best individual predictor is kept. The present work addresses the possibility that shared variance across indicators, namely for affect variability, may be better captured using latent models that also could better predict psychopathology. Methods: A 14-day EMA protocol was used to examine affect dynamics in 109 college-aged participants. Measures of psychopathology were collected on the first and last days. A minimum of 12 observations of the Positive and Negative Affect Schedule reports were needed for each participant. Measures of affect variability, granularity, and co-occurrence were derived. Results: Depression, anxiety, stress, and neuroticism were positively associated with latent negative affect variability and negatively associated with latent positive affect variability. Granularity and co-occurrence were not significant predictors. Importantly, latent factors were significantly stronger predictors of depression than within-person mean and standard deviations. Limitations: As with any latent variable study, the factorization is sample-specific and may have limited generalizability. Replication with a clinical sample and larger battery of psychopathology assessments is recommended. Conclusions: Latent factors coalesce the strengths of several EMA-derived indicators while maintaining statistical and construct validity. Clinical implications are discussed regarding short-burst daily affect assessments to track potential risk for depression onset.