Network-meta analysis made easy: detection of inconsistency using factorial analysis-of-variance models

Table 6 Studentized residuals and PRESS residuals

Design	Observation	Treatment	Model for heterogeneity
Design	Observation	Treatment	G.S.T fixed		G.S.T random		G.S.T dropped
			PRESS residual	Studentized residual	PRESS residual	Studentized residual	PRESS residual	Studentized residual
1	1	acar	0.0545	0.2443	0.0785	0.1453	0.0642	0.2925
	2	plac	-0.0545	-0.2443	-0.0785	-0.1453	-0.0642	-0.2925
2	3	acar	-0.0234	-0.1022	0.0619	0.1056	-0.0259	-0.1142
	4	SUal	0.0234	0.1022	-0.0619	-0.1056	0.0259	0.1142
3	5	benf
	6	plac	.	.	.	.	.	.
4	7	metf	0.0547	0.3026	-0.0781	-0.2282	-0.0814	-0.6783
	8	plac	-0.0547	-0.3026	0.0781	0.2282	0.0814	0.6783
5	9	acar	-0.0894	-0.2408	-0.1507	-0.2601	-0.1137	-0.3070
	10	metf	-0.0276	-0.0930	0.0036	0.0075	0.0060	0.0205
	11	plac	0.1359	0.3615	0.1193	0.2273	0.1057	0.2833
6	12	metf	0.6807	3.6726	0.6095	1.1614	0.6910	3.8755
	13	SUal	-0.6807	-3.6726	-0.6095	-1.1614	-0.6910	-3.8755
7	14	migl	.	.	.	.	.	.
	15	plac	.	.	.	.	.	.
8	16	piog	-0.4337	-2.5934	-0.2802	-0.5585	-0.3638	-2.2987
	17	plac	0.4337	2.5934	0.2802	0.5585	0.3638	2.2987
9	18	metf	-0.4719	-2.9147	-0.2927	-0.5779	-0.3467	-2.3246
	19	piog	0.4719	2.9147	0.2927	0.5779	0.3467	2.3246
10	20	piog	-0.1074	-0.5173	-0.0073	-0.0141	-0.0445	-0.2173
	21	rosi	0.1074	0.5173	0.0073	0.0141	0.0445	0.2173
11	22	plac	-0.2802	-1.9593	-0.2100	-0.6391	-0.3181	-2.4974
	23	rosi	0.2802	1.9593	0.2100	0.6391	0.3181	2.4974
12	24	metf	-0.1105	-0.5920	-0.0616	-0.1610	-0.0179	-0.1005
	25	rosi	0.1105	0.5920	0.0616	0.1610	0.0179	0.1005
13	26	rosi	-0.7077	-3.7022	-0.6733	-1.2693	-0.7424	-3.9701
	27	SUal	0.7077	3.7022	0.6733	1.2693	0.7424	3.9701
14	28	plac	.	.	.	.	.	.
	29	sita	.	.	.	.	.	.
15	30	plac	.	.	.	.	.	.
	31	vild	.	.	.	.	.	.

Residuals for diabetes example of Senn et al. [7] were obtained by fitting the model G + T to design × treatment means computed from model (2) with different assumptions regarding the effect for heterogeneity (G.S.T).

ISSN: 1471-2288