Comparison of different statistical models for the analysis of fracture events: findings from the Prevention of Falls Injury Trial (PreFIT)

Table 4 Goodness of fit statistic and likelihood test results for model comparisons

Regression Models	AIC	BIC	LR test^e
Poisson
Unadjusted	3875.85	3897.42	- -
Adjusted¹	3528.97	3579.11	- -
Mixed effects Poisson
Unadjusted	3866.23	3895.00	Chi-square value = 11.62^b
Adjusted^a	3527.10	3584.40	p-value < 0.001
Negative Binomial (NB)
Unadjusted	3652.17	3680.93	Chi-square value = 225.68^c
Adjusted^a	3363.62	3420.93	p-value < 0.001
Mixed effects NB
Unadjusted	3652.50	3688.45	Chi-square value = 1.67^d
Adjusted^a	3363.62	3420.93	p-value = 0.10

AIC Akaike Information Criterion, LR Likelihood ratio. ^aAdjusted for baseline covariates age, gender, deprivation score and log of baseline falls count, ^bLR test of mixed effects Poisson regression vs standard Poisson regression, ^cLR test of dispersion, ^dLR test of mixed effects NB regression vs standard negative binomial regression, and ^ethe LR tests results are for the unadjusted models

ISSN: 1471-2288