- Fixing a test that fails on r-universe.
- Fixed a bug where the
check_timing
wasn't taking the right arguments, and wasn't actually calculating what it should. - Added the
completeness
metric to the output ofici_kt
(and therefore toici_kendalltau
) (closes #23). - Added more tests (closes #22 for now).
- Fixed a bug where passing a two vector list is supposed to restrict the comparisons to just those provided, but instead did all possible pairwise comparisons between the two things and all others available.
- Refactored much of
ici_kendalltau
, making the code more consistent and easier to extend, as well as providing more informative error messages. Thanks to @njtierney for suggestions. - Also refactored
kt_fast
to be more consistent and use more functions internally. Note: now returns matrix or data.frame regardless of whether passing simply two vectors or a matrix input. - Added
alternative
andcontinuity
arguments to bothici_kendalltau
andkt_fast
. - Added
cor_fast
to allow running many iterations ofcor.test
on large matrix inputs if desired, with parallel processing to speed things up.
- makes
rank_order_data
take a sample class argument to enable splitting out by class.
- adds the
rank_order_data
function to help with visualizing the pattern of missingness with respect to the rank of the rows. - adds the
yeast_missing
dataset as a more realistic dataset for running thetest_left_censorship
andrank_order_data
.
- adds the function
test_left_censorship
to verify ifici_kendalltau
is appropriate to use on the data or not.
- Calculates correlation between columns of the matrix, not the rows.
kt_fast
now uses the same data.frame format for output asici_kendalltau
, but returns a matrix by default. The data.frame is useful when large amounts of comparisons are run.
- Handling parallel execution differently to avoid really large matrix issues on each core.
- Introduces the
return_matrix
parameter toici_kendalltau
that allows to return the results in the form of adata.frame
instead of lists of matrices.
- Provide the
kt_fast
function that handles missing orNA
values similarly to thestats::cor
function, but uses ourici_kt
fast function underneath.
- Made all error outputs the same length as the default output but containing
NA
. - Check for the case when one of the variables in
ici_kt
has all identical arguments, warns the user and returnsNA
.
- Added new argument
include_only
toici_kendalltau
that allows the user to define which of the pairwise correlations to actually do.
- switched theme.
- updated installation instructions to use r-universe.
- updating documentation and examples
- Switched
pairwise_completeness
to also use theglobal_na
parameter. - Oh yes, there is a function
pairwise_completeness
to enable scaling by the completeness between any two samples.
- Updated the API to use a single variable,
global_na
that defines all the values that should be set toNA
for the correlation calculation. - This is a big API change, so we bumped the version up to 0.1.
- Fixing a bug where if there are 55,000 elements in the vector, some of the match overflows from 32 bit to 64 bit, and results make no sense.
- Fixed a bug where instead of returning a two element vector, it only returned a zero length value. This probably only happened when one of the entries contained all NA values, or you tried to pass "vectors" with less than two entries.
- Fixing things that came up with R CMD check in the documentation. Should be all good now (I hope).
- Added tests!
- Removed a bunch of code that wasn't necessary because it was using the incorrect formula's, and we have the correct version in the
ici_kt_pairs
function for reference purposes.
- Fixed an issue where a warning would be issued if {furrr} was not installed.
- First release. It's been tested and used rather thoroughly by myself, but I'd say it's still rather alpha. Even though I use it all the time myself.
- Added a
NEWS.md
file to track changes to the package.