Accuracy of adjusted means against known values — prediction

In a simulation the true genotype values are known. This function measures how well an analysis recovered them: the correlation between the adjusted means and the true values.

Usage

prediction_accuracy(
  fit,
  truth,
  genotypes = NULL,
  method = c("pearson", "spearman")
)

Arguments

fit: An analysis object (from any analyze_*() function) or a data frame with genotype and adjusted_mean columns.
truth: Named numeric vector of true genotype values; names are the genotype names.
genotypes: Optional subset of genotypes to use (for example only the test entries, without the checks).
method: Correlation method: "pearson" (default) or "spearman" (rank correlation, i.e. how well the ranking is recovered).

Value

A single number between -1 and 1.

Examples

design <- augmented_prep_design(paste0("G", 1:30), c("A", "B"), block_size = 6, seed = 1)
trial <- design$design
set.seed(1)
truth <- setNames(rnorm(30, sd = 3), paste0("G", 1:30))
trial$yield <- 50 + truth[trial$all_entries] + rnorm(nrow(trial))
fit <- analyze_augmented_prep_design(trial, "yield")
prediction_accuracy(fit, truth)
#> [1] 0.9871109