Table 1.

Model diagnostics for checking whether an outcome or a predictor belongs in the COPE model. Left three columns: the Bayesian Wald statistic W_s in extended model A is used to test whether an outcome has a separate set of regression coefficients different from other outcomes. The p-value next to W_s comes from comparing W_s with χ₈², a χ²-distribution with 8 d.f. Anxiety is the only outcome whose Wald statistic is significant. We exclude anxiety from the COPE model. Right three columns: Wald statistics W_ϕ are from extended model B and are used to test whether a predictor has a different regression coefficient for each outcome. The p-value next to W_ϕ comes from comparing W_ϕ with a χ₇² distribution. Suicide attempt is the only predictor whose Wald statistic is significant.

outcome	Ws	p-value	predictor	Wϕ	p-value
depression	10.2	0.25	male	8.5	0.29
obsessive compulsive	9.8	0.28	age	4.2	0.76
BS11 (poor appetite)	5.5	0.70	food	8.1	0.33
BS13 (temper uncontrol)	10.3	0.24	finance	4.0	0.78
BS21 (people not friendly)	5.4	0.72	marijuana	6.2	0.52
BS31 (avoid frighten)	9.0	0.34	hard drugs	9.6	0.21
anxiety	15.6	0.05	AIDS symptoms	6.8	0.45
somatization	13.7	0.09	suicide attempt	14.7	0.04