Comparison of analysis methods and design choices for treatment-by-period interaction in unidirectional switch designs: a simulation study

Table 2 Type I error and power of GLMM and GEE at individual and aggregated level for testing treatment-by-period interactions with exchangeable correlation structure (\(N_h\) denotes the number of clusters allocated to switch pattern h)

Design	\(\sigma _c^2\)	Cluster allocation						Type I error				Power
Design	\(\sigma _c^2\)	\(N_0\)	\(N_1\)	\(N_2\)	\(N_3\)	\(N_4\)	\(N_5\)	GLMM Agg	GEE Agg	GLMM Ind	GEE Ind	GLMM Agg	GEE Agg	GLMM Ind	GEE Ind
PGD	0.2	24	0	0	0	0	24	0.053	0.050	0.052	0.050	0.994	0.981	0.994	0.981
PGD	0.5	24	0	0	0	0	24	0.040	0.044	0.039	0.042	0.992	0.917	0.991	0.914
PGD	0.8	24	0	0	0	0	24	0.060	0.056	0.060	0.055	0.984	0.842	0.984	0.829
DSDu	0.2	16	0	0	16	0	16	0.043	0.040	0.043	0.039	0.962	0.861	0.961	0.858
DSDu	0.5	16	0	0	16	0	16	0.040	0.037	0.039	0.037	0.948	0.712	0.948	0.704
DSDu	0.8	16	0	0	16	0	16	0.046	0.051	0.046	0.051	0.951	0.606	0.951	0.588
DSDo	0.2	14	0	0	24	0	10	0.049	0.035	0.050	0.034	0.913	0.724	0.911	0.716
DSDo	0.5	14	0	0	24	0	10	0.042	0.024	0.041	0.022	0.893	0.557	0.893	0.543
DSDo	0.8	14	0	0	24	0	10	0.043	0.037	0.042	0.038	0.857	0.451	0.854	0.445
USDu	0.2	8	8	8	8	8	8	0.047	0.036	0.047	0.036	0.858	0.659	0.858	0.653
USDu	0.5	8	8	8	8	8	8	0.032	0.037	0.032	0.037	0.841	0.511	0.840	0.510
USDu	0.8	8	8	8	8	8	8	0.059	0.046	0.058	0.046	0.827	0.405	0.825	0.402
USDo(1)	0.2	6	9	9	9	9	6	0.039	0.034	0.038	0.034	0.777	0.487	0.775	0.481
USDo(1)	0.5	6	9	9	9	9	6	0.032	0.022	0.031	0.022	0.737	0.365	0.733	0.363
USDo(1)	0.8	6	9	9	9	9	6	0.035	0.030	0.034	0.031	0.696	0.295	0.619	0.289
USDo(2)	0.2	5	10	9	9	10	5	0.036	0.029	0.036	0.029	0.700	0.415	0.698	0.408
USDo(2)	0.5	5	10	9	9	10	5	0.035	0.021	0.035	0.022	0.676	0.320	0.674	0.314
USDo(2)	0.8	5	10	9	9	10	5	0.036	0.034	0.034	0.035	0.644	0.255	0.642	0.251

ISSN: 1471-2288