Perform Prediction Cross-validation to determine best model — Prediction_cross

This function is used to calculate predictive powers of different models at different thresholds.

Usage

Prediction_cross_validation(
  rds.obj,
  k,
  threshold = 1,
  method = "GWAS",
  liabilities = rds.obj$FAM$Status
)

Arguments

rds.obj: A .rds file with an FBM.code256 and accompanying FAM and MAP tibbles.
k: Number of folds to be used in cross-validation. The number of rows in the FBM must be at least twice as large as k. Highly recommended to choose k to be at most ~1% of the number of rows, unless working with a very small dataset, as errors may occur.
threshold: Vector of significance levels to be used in thresholding. Default does not use thresholding.
method: Method to use for prediction. Possible methods are "GWAS", "GWAX", "LTFH". Default is "GWAS".
liabilities: Vector of liabilities used for prediction with "LTFH" method. If not specified, uses "GWAS" method instead.

Value

A list with 2 entries: a tibble with average and best scores for each threshold, and a data.frame with the best model, fitted values, residuals, best P-value and its R^2.