pcv.RdImplements the cross-validation described in Davison & Davenport (2002).
pcv(
formula,
data,
seed = NULL,
na.action = "na.fail",
family = "gaussian",
weights = NULL
)An object of class formula of the form response ~ terms.
An optional data frame, list or environment containing the variables in the model.
Should a seed be set? Function defaults to a random seed.
How should missing data be handled? Function defaults to failing if missing data are present.
A description of the error distribution and link function to be used in the model. See family.
An option vector of weights to be used in the fitting process.
An object of class critpat is returned, listing the f ollowing components:
R2.full, test of the null hypothesis that R2 = 0
R2.pat, test that the R2_pattern = 0
R2.level, test that the R2_level = 0
R2.full.lvl, test that the R2_full = R2_level = 0
R2.full.pat, test that the R2_full = R2_pattern = 0
The pcv function requires two arguments: criterion and predictor. The criterion corresonds to the dependent variable and the predictor corresponds to the matrix of predictor variables. The function performs the cross-validation technique described in Davison & Davenport (2002) and an object of class critpat is returned. There the following s3 generic functions are available: summary(),anova(), print(), and plot(). These functions provide a summary of the cross-validation (namely, R2); performs ANOVA of the R2 based on the split for the level, pattern, and overall; provide output similar to lm(); and plot the estimated parameters for the random split. Missing data are presently handled by specifying na.action = "na.omit", which performs listwise deletion and na.action = "na.fail", the default, which causes the function to fail. A seed may also be set for reproducibility by setting the seed.
Davison, M., & Davenport, E. (2002). Identifying criterion-related patterns of predictor scores using multiple regression. Psychological Methods, 7(4), 468-484. DOI: 10.1037/1082-989X.7.4.468.