Table 4.

Summary of multinomial regression block-entry ^a results testing the independent influence of meeting attendance class membership and sponsor class membership on abstinence class membership.

	High (vs. low) abstinence				Descending (vs. low) abstinence
	β	(se)	OR	p-val	β	(se)	OR	p-val
High (vs. low) attendance	1.36	(0.53)	3.9	.010	.23	(0.84)	1.3	.787
Descend. (vs. low) attendance	.85	(0.32)	2.3	.008	−.17	(0.59)	0.9	.782
Medium (vs. low) attendance	.71	(0.36)	2.0	.035	.32	(0.56)	1.4	.569
Descend. (vs. high) attendance	−.52	(0.56)	0.2	.360	−.39	(0.89)	0.6	.662
Medium (vs. high) attendance	−.66	(0.57)	0.2	.250	.09	(0.86)	.52	.918
High (vs. low) sponsor	1.95	(0.43)	7.0	<.001	2.48	(0.63)	12.0	<.001
Descend. (vs. low) sponsor	.68	(0.34)	2.0	.042	1.83	(0.54)	6.3	.001
High (vs. descend) sponsor	1.27	(0.48)	3.3	.009	.65	(0.63)	1.9	.303
Age	.02	(0.01)	>1.0	.041	.01	(0.03)	1.0	.805
Male (vs. female)	.36	(0.23)	1.4	.113	.90	(0.39)	2.5	.020
Agnostic/atheist/other vs. religious	−.63	(0.32)	0.5	.047	−.26	(0.55)	0.8	.639
Spiritual (vs. religious)	.03	(0.24)	1.0	.910	.03	(0.43)	1.0	.937
Gen. pop. sample (vs. treated)	−1.09	(0.31)	0.3	.001	−1.03	(0.67)	0.4	.122

High and the descending abstinence classes are contrasted to the low abstinence class (see table headings), that is, the low abstinence class was designated as the comparison group (coded 0) in the multinomial regression model. Contrasts between attendance classes within each abstinence contrast are noted in the rows. The same is true for the sponsor contrasts. Reference categories for the attendance and sponsor classes were changed to establish the various contracts displayed above; this has no bearing on p-values nor does it increase Type I errors. β=beta coefficient; se=standard error; OR=odds ratio.

^aBlock-1 (BIC=818; −2 log likelihood=706) and block-2 (BIC= 807; −2 log likelihood=671; Chi-Square=35.8, p<.001).