perm_rsq retrieves the \(R^2\) values from successful n_perm permutations
of a regression function.
perm_rsq_pvalue post-processes this vector to count how many of its entries
exceed the original \(R^2\) value. That fraction is a p-value for the null
hypothesis that the regression result is meaningless.
Arguments
- n_perm
A real number (positive integer). The number of permutations to try.
- FUN
An R function. Must return a result list with fields
convergence,min_eigenvalue,r_squaredTypicallyregression_greatcircle()or a similar regression function.- x
A real vector. The values of the independent variable. Assumed to be the first argument passed to
FUN.- ...
Other arguments passed to
FUN, afterx.- r_perm
vector of permutated \(R^2\) values
- r
original \(R^2\) value
Value
perm_rsq returns a a real vector. The maximum length is n_perms.
Often the length is less than n_perms, because the regression failed (as
signaled by error != 0 or min_eigenvalue <= 0).
perm_rsq_pvalue returns the p-value.
See also
Other geodesic-regression:
best_fit
Examples
set.seed(20250411)
data("gray_example")
# original regression
bestgc_clea <- regression_greatcircle(gray_example[1:8, ])
# permutation test
pr <- perm_rsq(100, FUN = regression_greatcircle, x = gray_example[1:8, ])
# p-value
perm_rsq_pvalue(pr, bestgc_clea)
#> [1] 0.03125