4.3.2. The Time of Measurement Model

To test the effect of the time of measurement, we entered T0, T4, and T5 measurements to the model (see 'time of measurement model' in Table 2). The reason for which we chose these three measurement points is that we were interested in possible long-term phase changes and not in day-to-day changes. Moreover, we wanted to keep the same number of points of measurement across most of the analyses performed. Once again, we set custom contrasts for the time of measurement model in which T0 was compared to T4, T0 to T5, and T4 to T5. The difference between the time of measurement model and the null model was statistically significant, and the AIC value for the time of measurement was lower (838.16) than that of the null model (849.35). It indicated that this model fits the collected data better than the null model.

The time of measurement model explained 3% of variance of the dependent variable (based on McFadden pseudo-R2). The outcomes of the Wald tests revealed that the difference between T0 and T4 and T0 and T5 for phase change from F1 to F2 were significant (see Table 3 for more details). More specifically, the change from T0 to T4 increased the odds of phase change from F1 to F2 by 1.37. In addition, the change from phase F1 to phase F2 was 1.48 odds higher on T5 when compared to T0. Similar results were obtained for change from phase F3 to phase F4. The significant predictors were contrasts between T0 and T4, and T0 and T5. Change from T0 to T4 increased the odds of phase change by 1.48, and change from T0 to T5 increased the odds by 1.86. The contrasts between the time of measurements did not predict the likelihood of changing from the phase F2 to the phase F3.


**Table 3.** Multinominal regression coefficients of the time of measurement model.

## 4.3.3. The Interaction of the Group and the Time of Measurement

We compared the model with an interaction term (time of measurement × group) to the time of the measurement model. They were not significantly different. In conclusion, the best fitting model was the one with the time of measurement as the predictor of the phase change. It suggests that mere participation in the study independent of the group was the best predictor of changes from phase 1 (pre-decisional stage) to phase 2 (pre-actional stage) and from the phase 3 (actional stage) to phase 4 (post-actional stage), see Table 2.
