ABSTRACT
This chapter demonstrates time series analysis with ARIMA models as a technique to model temporal patterns in LIWC variable scores across psychotherapy sessions. These patterns reveal subtle dynamics of therapist–client communication and can be used as a forecasting tool for prognostic purposes. The logic and characteristics of the Box-Jenkins method of time series analysis are first introduced, emphasizing the concepts of autocorrelation, components of time series data, and time series models as ‘structural signatures’ of discourse dynamics across time. The case study demonstrates the six steps of this method on two sample time series – a CBT therapist and client's analytical thinking scores across 40 sessions, with model validation using the first 37 sessions as training data. The results show that therapist and client analytical thinking fit contrasting ARIMA models that nevertheless have the same order (AR(1) versus MA(1)). Follow-up analysis of these structural signatures in context suggest the occurrence of ‘asynchrony in tandem’, which provide complementary support to the CBT clustering results from Chapter 3. A more general level of analysis that examines the ‘modelability’ of time series with respect to background reality will also be discussed. Time series analysis thus provides an additional tool for research, training, and self-evaluation.
