Table 2 shows the unadjusted means and SDs of the patient and partner target measure jealousy items during the baseline, therapy, and follow-up phases. The patient data were analysed using a time-series intervention analysis (Box & Tiao, 1975) with the patients’ daily self-ratings as the dependent variables. The first stage of this analysis involved identifying the best Autoregressive Integrated Moving Average Model (ARIMA; Box & Jenkins, 1976) for each time-series for each patient prior to the therapeutic intervention. Inspection of the autocorrelation functions (ACF), partial autocorrelation functions (PACF), and Box-Ljung statistics indicated that a simple ARIMA(1,0,0) model would be suitable in each case. Two-step functions, one for therapy and one for follow-up, were then entered as predictor dummy variables in the ARIMA(1,0,0) models for the entire time-series of each dependent variable. These were used to assess the impact of therapy and the additional impact of follow-up on the time-series of each dependent variable. Analysis of the residual ACF and PACF profiles from each model indicated that there was a satisfactory fit (i.e., no significant values at the various time lags) for the CBT patient jealousy target measure items but not for the CAT patient target measure jealousy items. Using the ACF and PACF profiles, an ARIMA(1,0,2) model was identified as more appropriate for the CAT patient. Re-estimation using this alternative model indicated a satisfactory fit for each of the CAT patient variables, and was therefore used for the CAT patient analysis.
For the CBT patient, the ARIMA models showed no significant effects on patient jealousy for therapy, T(173) =−1.56, ns, or follow-up, T(173) =−.22, ns, on hypervigilance for therapy, T(173) =−1.44, ns, or follow-up, T(173) = .16, ns, on disinhibition for therapy, T(173) =−.98, ns, or follow-up, T(173) =−1.53, ns, on anxiety for therapy, T(173) =−1.07, ns, or follow-up, T(173) =−1.25, ns, or on self-esteem for therapy, T(173) = 1.67, ns, or follow-up, T(173) = 1.12, ns. The CAT patient showed reduced jealousy during therapy, T(211) =−2.38, p < .05, and again during follow-up, T(211) =−2.48, p < .05, reduced hypervigilance during therapy, T(211) =−8.07, p < .01, and again during follow-up, T(211) =−2.77, p < .01, reduced anxiety during therapy, T(211) =−4.36, p < .01, but not during follow-up, T(211) =−1.27, ns, greater self-esteem during therapy, T(210) = 4.44, p < .01, but reduced self-esteem during follow-up (from the level during therapy), T(210) =−2.04, p < .05, and no significant effects on disinhibition for therapy, T(211) =−1.30, ns, or follow-up, T(211) =−1.23, ns
Figure 1 shows the day-by-day changes in each patient's jealousy during baseline, therapy, and follow-up phases. A number of methods have been developed to estimate the size of intervention effects in single-case experimental designs. Comparing three such methods, Manolov, Solanas, and Leiva (2010) concluded that the percentage of treatment data points exceeding the baseline median (PEM; Ma, 2006) is an effective method when autocorrelation or trend is present in the data. PEM was therefore used here. Cohen (1998) divided the evaluation of effect sizes into three parts in which sizes 0.20, 0.50, and 0.80 are labelled as slight, moderate, and strong effects, respectively. For CBT, the effect sizes were .71 for jealousy, .56 for hypervigilance, .53 for disinhibition, .76 for anxiety, and .74 for self-esteem target jealousy symptom items. For CAT, the effect sizes were .91 for jealousy, .80 for hypervigilance, .80 for disinhibition, .89 for anxiety, and .90 for self-esteem target jealousy symptom items. It is worth noting in the CAT data in Figure 2 that jealousy was largely extinguished by the mid-period of the treatment phase (i.e., patient and partner scoring 1 ‘not at all’ consistently on that item).