- Fixed faulty class check in
test_hypothesis
and updated documentation for clarity.
- Added
calc_cumsum
which provides additional NA handling when calculating cummulative sums.
- Fixed LazyData CRAN error.
- Added
dplyr
to suggests.
- The final tabulate function,
tabulate_at_risk
, has been removed after being internalized by utile.visuals
package.
- Unused
dplyr
dependency removed.
vctrs
dependency added.
- Fixed incorrect estimate rounding in <1 checks for
paste_median
and paste_mean
.
- Fixed issue where
test_hypothesis
would not accept logical stratification data.
- Added more reliable type checking and recycling to
calc_duration
.
- All functions now use a consistent x/y argument style.
- Added
paste()
& paste0()
. Base functions with added option for NA removal. Default behavior identical to base functions.
- Removed
tabulate_model()
. Feature parity has been provided in utile.tables::build_table()
.
- Pruned dependency list.
- Overhauled documentation.
calc_duration()
: Now supports the return of a lubridate::duration()
object when the units parameter is left unspecified.
- Improved NA value handling of
calc_duration()
.
paste_freq()
:
- Added support for tallying factors.
- Removed default rounding of numerators <1 to 0.
- Calculated percentages may now return as infinity.
paste_
functions: Fixed parameter naming [BREAKING CHANGE]
- Consolidated
test_
functions into test_hypothesis()
which has S3 methods for numeric, factor, and logical data. Numeric methods now support >2 strata. [BREAKING CHANGE]
- Added
calc_chunks()
& chunk_data_()
. Tools for breaking tibbles, data frames, and vectors into smaller, usable chunks of data.
tabulate_model()
: Formatted tables now combine the point estimate and confidence interval into a single column to be more consistent with the output of utile.tables::
functions.
paste_freq()
: Non-numeric data is now tallied and documentation has been updated.
paste_mean()
/paste_median()
/paste_efs()
: Added checks for numeric data.
calc_duration()
: Added support for multiple calculations against a single timepoint.
tabulate_model()
: Rethought the purpose of this function.
- Added
format=
option to return a human-readable, ready-to-export table.
- Added
tabulate_model.glm()
which only supports the ‘binomial’ family of glm functions for now.
- Added
tabulate_model.lm()
for linear regression model support.
- Removed
tabulate_model.survfit()
as other packages (i.e. ‘broom’) already provide this functionality.
- Refactored function design to make adding new methods easier.
tabulate_logit()
: Removed now that support has been added to tabulate_model()
.
tabulate_at_risk()
: Now returns the strata column as factor instead of character type.
paste_freq()
: Switched from an S3 object to a regular function.
- Fixed NA handling.
- Added support for vectorized data (i.e.
dplyr::mutate()
).
- Removed support for logical data. This seemed to be an edge case and was interfering with NA handling. May reconsider in future.
paste_median()
/paste_mean()
: Fixed logic for returning ‘<1’ if a statistic rounds to 0 and less.than.one=TRUE
.
calc_duration()
:
- Fixed NA handling.
- Added hard stops for data types and vectors lengths.
- Improved handling of vectorized data (i.e.
dplyr::mutate()
).
- Added logic to ignore timestamps if mixed data types given (i.e. start = POSIXt, end = Date).
- Removed rounding of negative numbers to 0.