Due to the relatively small sample size, the CI for borderline significant independent variables was inspected (ie, inspection of the lower limit CI in relation to the size of effect estimate) for indications of incorrect omission from final models. 36 All final models were deemed to have sufficient CI coverage, based on the events-per-variable ratio. 36 Using methods described by Cleves et al , 37 final models showed no violations of underlying assumptions of proportional hazards (eg, tests based on reestimation, interaction of analysis time with the independent variables and graphically through Schoenfeld residuals) and showed appropriate model fit.
← all excerpts
A longitudinal observational study of back pain incidence, risk factors and occupational physical activity in Swedish marine trainees.
1
—
—