Bivariate Ripley's K for Whole Slide Images (removed)

Removed in spatialTIME 2.0.0. Use bi_ripleys_k() instead, which now handles whole slide images directly. Calling this function raises an error.

Usage

bi_ripleys_k_WSI(
  mif,
  mnames,
  r_range = 0:100,
  edge_correction = "translation",
  num_permutations = 50,
  permute = FALSE,
  keep_permutation_distribution = FALSE,
  overwrite = FALSE,
  workers = 1,
  big = 1000,
  nlarge = 1000,
  xloc = NULL,
  yloc = NULL,
  ...
)

Arguments

mif, mnames, r_range, edge_correction, num_permutations, permute

Ignored.

keep_permutation_distribution, overwrite, workers, big, nlarge, xloc, yloc

Ignored.

...

Ignored.

Value

Nothing. Always throws an error directing you to bi_ripleys_k().

Details

bi_ripleys_k_WSI() existed because bi_ripleys_k() used to build an n-by-n distance matrix, which is impossible at whole-slide scale, and so a separate tiled implementation was maintained alongside it. The two functions were otherwise near-duplicates, and the tiled one traded accuracy for memory: above a cell-count threshold it replaced the requested edge correction with "none", its tiled branch hard-coded translation regardless of the edge_correction argument, and it silently returned garbage for edge_correction = "isotropic" because that branch of its inner loop returned NULL.

As of 2.0.0 that trade-off is gone. bi_ripleys_k() only ever materialises the cell pairs closer than max(r_range), so whole-slide images are handled directly – at 200,000 cells roughly 75 MB rather than 319 GB – with the requested edge correction always honoured and results equal to spatstat.explore::Kcross() to floating-point precision.

Migration is a rename. big and nlarge have no equivalent because they no longer affect the statistic; bi_ripleys_k() takes a big argument that bounds peak memory only.

# before
bi_ripleys_k_WSI(mif, mnames, r_range = 0:100, big = 1000, nlarge = 1000)

# after
bi_ripleys_k(mif, mnames, r_range = 0:100)

See also

bi_ripleys_k()