ARIMA modelling of chronospsychometric self-evaluation of drive and mood.

Abstract

The expanding study of biological rhythms requires the use of refined methods of time series analysis. We propose the use of ARIMA (Autoregressive Integrated Moving Averages) model, a powerful statistical tool of relatively recent development. A group of 5 patients with Affective Disorders (2 bipolars, 3 unipolars) and 1 patient with Adjustment Disorder self-assessed their AS every 8 hours for about a month. The affective state (AS) was estimated for 4 indicators: the two main constructs (Mood and Drive) of the segmented Visual Scale ESTA III and two bipolar items (Anxiety and Drowsiness). Mood and Drive are continuous variables, while Anxiety and Drowsiness are ordinal ones. Strictly speaking, ARIMA modelling is not valid with ordinal data. However, comparison of models of the two kinds of variables reveals no significant differences. This points out to a certain robustness of the method. Most of the series were non-stationary but could be transformed taking no more than two differences. The models made a very good fit of the data. Statistically significant coefficients on different lags may indicate the presence of circadian and infradian periodicities in the series. Further applications of ARIMA models to biological and psychological rhythmometry may be quite useful.