Modelling count, bounded and skewed continuous outcomes in physical activity research: beyond linear regression models

Table 3 Estimated associations of age, sex and NCMC with MVPA from different regression models for the ALECS study of physical activity in Hong Kong older adults (n = 402)

	LM	sqrt-LM	log-LM	GLM-gamma	GLM-IG
Intercept
Estimate	115.567	14.835	8.754	7.742	8.548
SE	12.438	1.202	0.673	0.521	0.537
p-value	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001
95% CI	91.10, 140.00	12.47, 17.20	7.43, 10.08	6.66, 8.81	7.39, 9.69
Age (years)
Estimate	-1.263	-0.145	-0.085	-0.065	-0.076
SE	0.166	0.016	0.009	0.007	0.007
p-value	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001
95% CI	-1.59, -0.94	-0.18, -0.11	-0.10, -0.07	-0.08, -0.05	-0.09, -0.06
Sex (Female reference)
Estimate	19.593	1.885	0.899	0.754	0.816
SE	2.186	0.211	0.118	0.092	0.123
p-value	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001
95% CI	15.29, 23.89	1.47, 2.30	0.67, 1.13	0.58, 0.94	0.59, 1.08
NCMC
Estimate	-0.177	0.010	0.029	0.012	0.027
SE	0.505	0.049	0.027	0.021	0.022
p-value	0.727	0.841	0.289	0.559	0.219
95% CI	-1.17, 0.82	-0.09, 0.11	-0.02, 0.08	-0.03, 0.05	-0.01, 0.07
Adjusted-AIC	3563	3321	3377	3320	3599

LM: linear regression model fit on untransformed MVPA; sqrt-LM: linear regression model fit on square-root transformed MVPA; log-LM: linear regression model fit on log transformed MVPA; GLM: generalized linear model; IG: inverse Gaussian; SE: standard error; CI: confidence interval; AIC: Akaike Information Criterion

ISSN: 1479-5868