Table 3.

Summary of estimation results in WGEE and doubly robust GEE with K = 100 and T = 3.

Type		Parameters	True	Correct dropout model			Mis-specified dropout model
				Bias	SE	MSE	Bias	SE	MSE
	wgee	β0	−0.5	0.0136	0.1113	0.0122	−0.0162	0.1094	0.0122
		β1	0.5	−0.0087	0.1658	0.0271	−0.0180	0.1623	0.0266
Continuous
	drgee	β0	−0.5	−0.0198	0.0699	0.0123	−0.0093	0.0690	0.0118
		β1	0.5	−0.0083	0.0925	0.0220	−0.0106	0.0909	0.0218
	wgee	β0	−0.5	−0.0066	0.2435	0.0664	−0.0432	0.2414	0.0665
		β1	0.5	0.0178	0.3602	0.1309	0.0048	0.3560	0.1293
Binary
	drgee	β0	−0.5	−0.0380	0.1330	0.0615	−0.0283	0.1329	0.0609
		B1	0.5	0.0124	0.1769	0.1111	0.0110	0.1764	0.1110
	wgee	β0	−0.5	−0.0012	0.1483	0.0301	−0.0316	0.1451	0.0278
		β1	0.5	−0.0121	0.2007	0.0511	−0.0231	0.1966	0.0463
Count
	drgee	β0	−0.5	−0.0298	0.0779	0.0240	−0.0238	0.0776	0.0239
		B1	0.5	−0.0057	0.0956	0.0348	−0.0071	0.0950	0.0350

Notes: The missing proportion is 28% for the data with continuous outcome, 32% for the data with binary outcomes and 35% for the data with count outcomes. The true exchangeable correlation structure is used for both models. Bias is the difference between the mean of the parameter estimates and the true value; SE is the mean of the standard error estimates and MSE is the mean square error of the parameter estimates.