Both logistic regression and Bayesian models (to minimize type-1 errors) were used, adjusted for potential confounders.
The time-stratified control sampling approach was used, adjusted for particulate matter (PM10) concentrations, time trends and meteorological influences.
Logistic regression analyses were used, adjusted by gender.
For analyses, repeated-measures analysis of variance was used, adjusted for covariates.
Unconditional logistic regression in successive models was used, adjusted for socio-demographic variables and other confounders.
In addition, specific diagnostic algorithms can be used, adjusted for age or developmental level [ 3, 15].
To adjust for confounding factors, a multiple logistic regression analysis was used, adjusted for age, sex, and comorbidity.
The cumulative average intake level involving information from 1980 , 1984 , 1986 , 1990 , 199498, 2002 and 2006 was used, adjusted for total energy intake.
Published mortality rates per 100,000 from the National Center for Health Statistics [ 11, 12] were used adjusted to the 2000 US Standard Million population.
23 Poisson regression was used, adjusting for propensity score deciles.
