BLiP when the set of locations is continuous, e.g., when working with image data.

Usage

BLiP_cts(
  locs,
  grid_sizes = DEFAULT_GRID_SIZES,
  weight_fn = inverse_radius,
  extra_centers = NULL,
  max_pep = 0.25,
  shape = "circle",
  min_blip_size = 5000,
  verbose = T,
  ...
)

Arguments

locs: A (N, num_disc, d)-dimensional array. Here, N is the number of samples from the posterior, d is the number of dimensions of the space, and each point corresponds to a signal in a particular posterior sample.
grid_sizes: List of grid sizes to split up the locations. The grid size is inversely proportional to the distance between lattice points.
weight_fn: Weight function which maps candidate groups to weights. Defaults to inverse radius.
extra_centers: An (ncenters, d)-dimensional matrix. At each resolution, candidate groups will be computed with centers at this location.
max_pep: The maximum allowable PEP for output candidate groups. Defaults to 0.25.
shape: One of ('circle', 'square').
min_blip_size: Combines connected components so all subproblems are at least this size.
verbose: If T, gives occasional status updates. Defaults to T.
...: Other arguments to the underlying BliP call, such as the error rate or nominal level.