This function conducts k-fold cross-validation for springer and returns the optimal values of the tuning parameters.
cv.springer(
clin = NULL,
e,
g,
y,
beta0,
lambda1,
lambda2,
nfolds,
func,
corr,
structure,
maxits = 30,
tol = 0.001
)an object of class "cv.springer" is returned, with is a list with components below:
the optimal \(\lambda_{1}\).
the optimal \(\lambda_{2}\).
a matrix of clinical covariates. The default value is NULL. Whether to include the clinical covariates is decided by user.
a matrix of environment factors.
a matrix of genetic factors.
the longitudinal response.
the initial value for the coefficient vector.
a user-supplied sequence of \(\lambda_{1}\) values, which serves as a tuning parameter for the individual-level penalty.
a user-supplied sequence of \(\lambda_{2}\) values, which serves as a tuning parameter for the group-level penalty.
the number of folds for cross-validation.
the framework to obtain the score equation. Two choices are available: "GEE" and "QIF".
the working correlation structure adopted in the estimation algorithm. The springer provides three choices for the working correlation structure: exchangeable, AR-1,and independence.
Three choices are available for structured variable selection. "bilevel" for sparse-group selection on both group-level and individual-level. "group" for selection on group-level only. "individual" for selection on individual-level only.
the maximum number of iterations that is used in the estimation algorithm. The default value is 30.
The tolerance level. Coefficients with absolute values that are smaller than the tolerance level will be set to zero. The adhoc value can be chosen as 0.001.
For bi-level sparse group selection, cv.springer returns two optimal tuning parameters, \(\lambda_{1}\) and \(\lambda_{2}\); for group-level selection, this function returns the optimal \(\lambda_{2}\) with \(\lambda_{1}\)=0; for individual-level selection, this function returns the optimal \(\lambda_{1}\) with \(\lambda_{2}\)=0.