infoCriterion function

Function that calculates cross-validation selection criteria

infoCriterion(ynew, pred, family, type, size = NULL, npar = 0)

Arguments

• ynew: data matrix corresponding to the observations used as test sample.
• pred: predicted value of the linear predictor obtained from Xnew and the estimated parameters.
• family: a vector of the same length as the number of responses containing characters identifying the distribution families of the dependent variables. "bernoulli", "binomial", "poisson" or "gaussian" are allowed.
• type: information criterion used. Likelihood, aic, bic, aicc or Mean Square Prediction Error (mspe) are defined. Area Under ROC Curve (auc) also defined for Bernoulli cases only.
• size: describes the number of trials for the binomial dependent variables. A (number of statistical units * number of binomial dependent variables) matrix is expected.
• npar: number of parameters used for penalisation.

Returns

a matrix containing the criterion value for each dependent variable (row) and each number of components (column).