. 2019 May 29;10:1186. doi: 10.3389/fpsyg.2019.01186

Table 2.

Model fit, parameter estimates and corresponding standard errors of the fixed and random effects of the five multilevel models.

	Baseline		Emotion				Cognitive processes
	M0		ME1		ME2		MCP1		MCP2
Fixed	23.41 (0.792)	^*	22.83 (1.516)	^**	22.61 (1.397)	^**	21.62 (1.588)	^**	21.69 (1.447)	^**
Intercept	-7.99 (0.858)	^*	0.02 (0.035)		0.03 (0.045)		0.04 (0.027)		0.08 (0.057)
Post-indicator			-5.60 (1.735)	^*	-6.33 (1.649)	^**	-5.33 (1.820)	^*	-5.50 (1.663)	^**
Variable interaction			-0.06 (0.039)		-0.08 (0.066)		-0.05 (0.033)		-0.12 (0.068)
Random
$σ_{e}^{2}$	35.74		35.37		35.80		35.46		35.66
τ	25.06		24.89		25.21		25.21		24.72
Model fit
deviance	1327.38		1323.48		1325.89		1324.58		1324.24
AIC	1335.38		1335.48		1337.89		1336.58		1336.24
BIC	1348.45		1355.09		1357.50		1356.19		1355.85
LogLik	-663.69		-661.74		-662.95		-662.29		-662.12
χ²			3.89		1.48		2.80		3.13
χ² df			2		2		2		2
Effect size
$Ω_{0}^{2}$ df	0.67		0.68		0.67		0.68		0.67

The mean of the text-variable, indicated by “variable” in the table, changes between the five models: in ME1 it is the number of positive words, in ME2 it is the number of negative words, in MCP1 it is the number of insight words, and in MCP2 it is the number of cause words. The “interaction” variable is the interaction between the text variable and the post-therapeutic indicator (“post. indi.”).

Coefficients (and standard errors).

p < 0.01.

^**

p < 0.001.