Table 2.
Causal estimands for the effect of sessions and alliance
| Estimate | s.e. | p | 95% CI | |
|---|---|---|---|---|
| (a) Evaluated when best alliance = 0 | ||||
| (1) SEM ignoring participants without data on both sessions and alliance | ||||
| βS (effect of sessions at best alliance) | −2.66 | 0.76 | 0.000 | −4.15 to −1.17 |
| βSA (the interaction) | −1.44 | 0.50 | 0.004 | −2.42 to −0.47 |
| (2) SEM FIML including all participants’ data | ||||
| βS (effect of sessions at best alliance) | −2.91 | 1.02 | 0.005 | −4.91 to −0.90 |
| βSA (the interaction) | −1.52 | 0.63 | 0.016 | −2.76 to −0.29 |
| (b) Evaluated when worst alliance = 0 | ||||
| (1) SEM ignoring participants without data on both sessions and alliance | ||||
| βS (effect of sessions at worst alliance) | 7.45 | 2.76 | 0.007 | 2.04 to 12.87 |
| βSA (the interaction) | −1.45 | 0.50 | 0.004 | −2.42 to −0.47 |
| (2) SEM FIML including all participants’ data | ||||
| βS (effect of sessions at worst alliance) | 7.74 | 3.42 | 0.024 | 1.03 to 14.45 |
| βSA (the interaction) | −1.52 | 0.63 | 0.016 | −2.76 to −0.29 |
CI, Confidence interval; SEM FIML, structural equation modelling with full information maximum likelihood.