Source code is now maintained on GitHub by authors
HiClimR 1.1.5
Updating description, URL, and citation info
HiClimR 1.1.4
Addresses an issue for zero-length mask vector:
Error in -mask : invalid argument to unary operator
this error was introduced in v1.1.2+ after fixing the data-mean
bug
HiClimR 1.1.3
The user manual is revised
lonSkip and latSkip renamed to
lonStep and latStep, respectively
Minor bug fixes
HiClimR 1.1.2
A bug has been fixed where data mean is added to centered data if
standardize = FALSE
objective tree cut and data component are now corrected
to match input parameters especially when clustering of raw
data
centered data was used in previous versions
HiClimR 1.1.1
Minor bug fixes and memory optimizations especially for the
geographic masking function geogMask
The limit for data size has been removed (use with caution)
A logical parameter InDispute is added to
geogMask function to optionally consider areas in dispute
for geographic masking by country
HiClimR 1.1.0
Code cleanup and bug fixes
An issue with fastCor function that degrades its
performance on 32-bit machines has been fixed
A significant performance improvement can only be achieved when
building R on 64-bit machines with an optimized BLAS
library, such as ATLAS, OpenBLAS, or the
commercial Intel MKL
The citation info has been updated to reflect the current status of
the technical paper
HiClimR 1.0.9
Minor changes and fixes for CRAN
For memory considerations,
smaller test case with 1 degree resolution instead of 0.5
degree
the resolution option (res parameter) in geographic
masking is removed
Mask data is only available in 0.1 degree (~10 km) resolution
LazyLoad and LazyData are enabled in the
description file
The worldMask and TestCase data are
converted to lists to avoid conflicts of variable names
(lon, lat, info, and
mask) with lazy loading
HiClimR 1.0.8
Code cleanup and bug fixes
Region maps are unified for both gridded and ungridded data
HiClimR 1.0.7
Hybrid hierarchical clustering feature that utilizes the pros of the
available methods
especially the better overall homogeneity in Ward’s method and the
separation and objective tree cut of the regional linkage method.
The logical parameter hybrid is added to enable a
second clustering step
using regional linkage for reconstructing the upper
part of the tree at a cut
defined by kH (number of clusters to restart with using
the regional linkage method)
If kH = NULL, the tree will be reconstructed for the
upper part with the first merging cost larger than the mean merging cost
for the entire tree
merging cost is the loss of overall homogeneity at each merging
step
If hybrid clustering is requested, the updated upper-part of the
tree will be used for cluster validation.
HiClimR 1.0.6
Code cleanup and bug fixes
HiClimR 1.0.5
Code cleanup and bug fixes
Adds support to generate region maps for ungridded data
HiClimR 1.0.4
Code cleanup and bug fixes
The coarseR function is called inside the core
HiClimR function
Adds coords component to the output tree for the
longitude and latitude coordinates
they may be changed by coarsening
validClimR function does not require lon
and lat arguments
they are now available in the output tree (coords
component)
HiClimR 1.0.3
Code cleanup and bug fixes
One main/wrapper function HiClimR internally calls all
other functions
Unified component names for all functions
Objective tree cut is supported only for the regional
linkage method
Otherwise, the number of clusters k should be
specified
The new clustering method has been renamed from HiClimR
to regional linkage method
HiClimR 1.0.2
Code cleanup and bug fixes.
adds a new feature that to return the preprocessed data used for
clustering, by a logical argument retData.
the data will be returned in a componentdata of the
output tree
this can be used to utilize HiCLimR preprocessing
options for further analysis
Ordered regions vector for the selected number of clusters are now
returned in the region component
length equals the number of spatial elements N
HiClimR 1.0.1
Code cleanup and bug fixes
Adds a new feature in validCLimR that enables users to
exclude very small clusters from validation indices
interCor, intraCor, diffCor, and
statSum, by setting a value for the minimum cluster size
(minSize > 1)
the excluded clusters can be identified from the output of
validClimR in clustFlag component, which takes
a value of 1 for valid clusters or 0 for
excluded clusters
in HiClimR (currently, regional linkage)
method, noisy spatial elements (or stations) are isolated in very
small-size clusters or individuals since they do not correlate well with
any other elements
this should be followed by a quality control step
Adds coarseR function for coarsening spatial resolution
of the input matrix x
HiClimR 1.0.0
Initial version of HiClimR package that modifies
hclust function in stats library
Adds a new clustering method to the set of available methods
The new method is explained in the context of a spatiotemporal
problem, in which N spatial elements (e.g., stations) are
divided into k regions, given that each element has
observations (or timeseries) of length M
minimizes the inter-regional correlation between region means
modifies average update formulae by incorporating the
standard deviation of the mean of the merged region
a function of the correlation between the individual regions, and
their standard deviations before merging
equals the average of their standard deviations if and only if the
correlation between the two merged regions is 100%.
in this special case, the new method is reduced to the classic
average linkage clustering method
Several features are included to facilitate spatiotemporal analysis
applications:
options for preprocessing and postprocessing
efficient code execution for large datasets.
cluster validation function validClimR
implements an objective tree cut to find an optimal number of
clusters