LoginSign Up
← Back to all sparks
S

spatstat.random

ANALYTICS
Velocity2.5

Sub-package of spatstat containing functions for random generation

github.com/spatstat/spatstat.random ↗

spatstat's simulation engine pushes point process generation into three dimensions

spatial-statisticspoint-processessimulationr-packagethree-dimensional
◆Current state
spatstat.random generates random point patterns and simulates point process models for the spatstat family. Its recent releases have moved along two lines at once: filling out three-dimensional simulation, and adding conditional simulation to the established cluster process generators. 3.5-1 is a narrow follow-up adding a random Dirichlet-Voronoi tessellation without edge effects.
◆Where it's heading
The clearest arc is dimensional. 3.5-0 carried inhomogeneous Poisson processes, non-uniform random points and Simple Sequential Inhibition into 3D in a single release, and the sibling geometry package followed two months later with more capabilities for three-dimensional point patterns. Alongside that, the generators have been gaining theoretical range — Gaussian random fields in 3.4-4, a new class of theoretical cluster process models and random diffusion in 3.5-0 — while earlier releases concentrated on conditional simulation and efficiency in the existing 2D routines.
◆Prediction
Expect the 3D work to continue propagating into the model-fitting and geometry packages before spatstat.random adds another dimension-independent generator, since the 3D features here have already begun appearing downstream. The entries do not indicate which estimator gets 3D support next.

◆Recent moves

  1. 20d ago

    Edge-effect-free random Dirichlet-Voronoi tessellation

    A single-feature release adding a random Dirichlet-Voronoi tessellation generated without edge effects. Edge effects are the standard bias in tessellation simulation, so removing them matters more than the one-line entry suggests, but this is a narrow addition next to the 3D work in 3.5-0.

    View source ↗
  2. 2mo ago

    Three-dimensional point process simulation arrives

    ⚡ SPARK

    The release where three-dimensional simulation stops being a gap: inhomogeneous Poisson processes, non-uniform random points and Simple Sequential Inhibition all arrive in 3D at once, alongside random diffusion, random unrounding, a truncated Poisson distribution and a new class of theoretical cluster process models. The sibling geometry package extends the same 3D work two releases later.

    View source ↗
  3. 6mo ago

    Gaussian random field generation added

    Adds generation of Gaussian random fields, plus minor improvements and bug fixes. Gaussian random fields underpin log-Gaussian Cox process simulation, so this sits on the same widening of theoretical generators that 3.5-0 continued.

    View source ↗
  4. 10mo ago

    runifdisc efficiency and fixed-count simulation options

    Efficiency improvements to `runifdisc` and more options when simulating a fixed number of random points. Internal tuning of existing generators rather than new simulation capability.

    View source ↗
  5. 1y ago

    Conditional simulation for the cluster process generators

    Conditional simulation arrives across the cluster process generators — `rLGCP`, `rThomas`, `rMatClust`, `rCauchy` and `rVarGamma`. Simulating conditionally on an observed pattern is what these generators need to support model checking rather than only forward simulation.

    View source ↗
  6. 1y ago

    Faster rpoispp for tessellation-defined intensity

    An improved algorithm for `rpoispp` when the intensity `lambda` is defined by a tessellation. A performance path for one specific intensity representation.

    View source ↗