Skip to contents

Changelog

Source: NEWS.md

sfhotspot 1.1.1

  • Package tests now streamlined to reducing CRAN testing time (#98).

sfhotspot 1.1.0

  • New function hotspot_dbscan() to identify clusters using the DBSCAN algorithm, as implemented in the dbscan package. Its min_pts and eps parameters are selected automatically by default from the number of points and nearest-neighbour distances, respectively. Also added corresponding autoplot() and autolayer() methods.
  • New function hotspot_isoband() to generalise values in regular square hotspot grids into tidy SF isobands with appropriate visual representation (#81).
  • New function hotspot_map() to produce quick maps of the objects produced by functions in the hotspot_*() family, as well as of generic sf objects. By default, maps produced by hotspot_map() include a base map while those produced using autoplot() methods do not.
  • New function hotspot_layer() provides a student-friendly wrapper around ggplot2::autolayer() for adding sfhotspot results to custom ggplot2 maps.
  • hotspot_gistar() and hotspot_classify() now use the default Holm method when p_adjust_method = NULL. P-values are adjusted once using spdep::p.adjustSP(); hotspot_classify() no longer applies a second adjustment across time periods. Use p_adjust_method = "none" to retain unadjusted p-values (#94).
  • Added autoplot() and autolayer() methods for hotspot_gistar() and hotspot_dual_kde() results, and completed the plotting support for hotspot_classify() results. Plot scales and legends now reflect the semantics of each result, including the comparison method used for dual KDE and significance/sign filtering for Gi* results. A new website plotting article demonstrates each distinct behaviour (#85).
  • hotspot_clip() can now clip any type of geometry, not just point data (#65, #78) and produces warnings if clipping changes the geometry type of any features in a way that might create downstream problems.
  • hotspot_clip() now preserves the specialised hspt_* class of results produced by other package functions (#71).
  • hotspot_gistar() results now have the specialised class hspt_g, allowing downstream functions to identify gistar as their primary value column (#82).
  • Existing plotting methods now use transparent missing values, weighted counts when present, sequential scales anchored at zero, and balanced diverging scales for signed values (#85, #86).
  • hotspot_clip() no longer reports that zero rows were removed when all input features fall within the clipping boundary (#66).
  • hotspot_clip() no longer suppresses warnings produced by sf::st_intersection() other than the constant “attribute variables are assumed to be spatially constant throughout all geometries” (since this is rarely relevant).
  • hotspot_gistar() now calculates KDE values for longitude/latitude data by automatically transforming them to a projected co-ordinate reference system and transforming the results back afterwards (#68).
  • hotspot_dual_kde() now calculates valid KDE values for longitude/latitude data while using a common automatically selected bandwidth for both layers, and checks that both point layers overlap the analysis grid (#67).
  • hotspot_gistar() now uses weighted counts to calculate Gi* statistics and p-values when weights are supplied, and autoplot()/autolayer() plot use weighted counts when they are present (#86).
  • Automatic CRS transformation using st_transform_auto() in KDE-related functions now correctly restores the original geographic CRS rather than converting to EPSG:4326 even when another geographic CRS was originally used (#89).
  • Automatic CRS transformation using st_transform_auto() in hotspot_kde() and hotspot_gistar() now ensures the same CRS is selected for both data and grid (#90).

sfhotspot 1.0.0

CRAN release: 2025-07-29

  • New function hotspot_clip() added to extract points from an SF object inside the boundary of a polygon (#57).
  • By default lat/lon data is now transformed to use a projected co-ordinate reference system before kernel density estimation and then transformed back afterwards (previously trying to run KDE on lat/lon data resulted in an error). Datasets can also be automatically transformed to/from lat/lon using st_transform_auto() (#48).
  • Functions now return useful error if provided with an empty dataset (#58).
  • Adjusted bandwidth now reported accurately (#56).
  • Removed dependency on the rmarkdown package in favour of using the quarto package to generate vignettes.

sfhotspot 0.9.2

sfhotspot 0.9.1

CRAN release: 2025-02-19

  • hotspot_grid() handles non-multipolygon input geometries (#46).
  • hotspot_dual_kde(), hotspot_gistar() and hotspot_kde() now warn if KDE bandwidth is smaller than cell size (#29).
  • count_points_in_polygons() (which is used internally to count points in all the hotspot_*() family of functions) now respects quiet = TRUE (#52).
  • Conditions now formatted with the cli package and following the Tidyverse Style Guide for errors (#47).

sfhotspot 0.9.0

CRAN release: 2025-02-10

  • count_points_in_polygons() now passes through columns in the original dataset, which makes hotspot_count() more useful (#41).
  • hotspot_grid() if provided with polygons now bases the grid on the boundary of the polygons rather than the convex hull of the boundary (#42).
  • New dataset memphis_precincts showing Memphis Police Department precincts, which is required to test the new functionality of hotspot_grid().
  • hotspot_gistar() now extracts nearest neighbour distance from provided grid and does not wrongly rely on (and report) an automatically generated cell size (#38).
  • Warnings about grids containing very large numbers of cells is now printed before the cells are created, helping explain why code may be running slower than expected (#33).
  • Improved error message produced when point data and provided grid do not overlap (#39).
  • Suppressed progress bar previously included in README (#36).

sfhotspot 0.8.0

CRAN release: 2023-09-19

  • All functions can now handle SF objects in which the geometry column has a name other than geometry (#30).
  • hotspot_kde() progress bar is now suppressed by quiet = TRUE (#25).
  • When KDE bandwidth is selected automatically based on a rule of thumb, the reported bandwidth now takes into account the value of bandwidth_adjust (#32).

sfhotspot 0.7.1

CRAN release: 2023-02-10

  • Added warning if automatically created grids contain more than 100,000 cells, since this can cause other functions in the package to run very slowly (#24).
  • Fixed bug in which cell size could be incorrectly rounded to zero (#26).
  • Removed covr dependency.

sfhotspot 0.7.0

CRAN release: 2022-09-11

sfhotspot 0.6.1

  • Fixed bug in which the message produced when setting bandwidth automatically referred to the unadjusted rather than adjusted bandwidth (#22).

sfhotspot 0.6.0

  • New function hotspot_dual_kde() for estimating different relations between the density of two point layers (#1).
  • New dataset memphis_population showing the 2020 population of the centroids of census blocks in Memphis, TN.

sfhotspot 0.5.0

sfhotspot 0.4.0

  • Users can now provide their own grid using the grid argument to the hotspot_*() family of functions.
  • New function hotspot_grid() added so users can create a rectangular or hexagonal grid separately from counting points, calculating KDE values, etc. This may be useful to use the same grid for different datasets covering a similar area.

sfhotspot 0.3.1

  • Fixed bug where ... arguments were not passed on to SpatialKDE::kde() as specified in the documentation.

sfhotspot 0.3.0

sfhotspot 0.2.2

  • Added checking for empty geometries and co-ordinates on Null Island (or equivalent zero co-ordinates in other co-ordinate systems). Empty geometries produce an error and zero co-ordinates produce a warning.

sfhotspot 0.2.1

  • Fixed a bug where a grid of cells could not be created for an SF object with no CRS defined (and improved the error message when trying to calculate KDE values in this circumstance).
  • Fixed a bug where the grid returned by the hotspot_*() family of functions was not clipped to the convex hull of the data, as specified in the documentation.

sfhotspot 0.2.0

CRAN release: 2022-02-14

sfhotspot 0.1.1

  • Fixed a problem with some of the examples in the documentation taking too long to run, which caused them to fail CRAN checks.

sfhotspot 0.1.0