The multivariate methods described and used herein rely on linear algebraic operations and are transparent in contrast to neural nets (for example), which are often seen as black boxes.
We used a multivariate error correction method described by Wood et al (19) in the Fibrinogen Studies Collaboration.
Population attributable risks (PAR) were derived from multivariate ORs, using the method described by Bruzzi et al (1985) and Mezzetti et al (1996).
Sensitivity, specificity and accuracy were then modeled by use of multivariate logistic regression estimated through the method described in detail by Smith-Bindman et al [ 15].
Multivariate binary correlated data was generated using the method described in Park et al. [ 29].
For the spline method, we fitted multivariate second stage regression models based on the method described by Berkey et al. (1998).
The cluster-averaged residual for both treatment groups were compared using the method described, where is the cluster-averaged residual from a multivariate regression, used in place of the cluster-averaged mean in the method described above.
There is no method described in the literature for an exact power calculation for multivariate statistical methods with event times as the endpoint and equivalence as the main question of the analysis.
Once the final multivariate model was determined, the logistic model was converted to absolute risk using the method described in the Appendix.
Multivariate methods were described by Garner and Robertson, van der Voet, Olivieri et al., and Boqué et al. [ 30– 30].
