- Fix bug in Kalman Smoother
- Tidy up Kalman Smoother documentation to correct notation
- Fix memory leak in fks identified by CRAN valgrind tests
- Fix building on MAC arm64 (M1) platform
2021-09-28 Paul Smith
- Incorporated Kalman Smoother Code by Danial McDonald
- Outputs from the
fkf function altered for use by the
smoother. This includes a change in the output ordering. Code that
references by name (e.g.
out$at} should work as expected.
Code that references by location (e.g.
- Implimentation of USE_FC_LEN_T on calls to BLAS and LAPACK
2020-06-03 Paul Smith
- Fixed ASAN error caused by error in Vignette example
- Depreciated check.input option to fkf so dimensions always checked -
the reason for ongoing ASAN issues since v0.1.2.
2020-06-01 Paul Smith
- Tidying NAMESPACE to fix S4 class export warning with FKF v0.1.5 on
all linux R-devel
- Converted Vignettes to use knitr
2018-07-16 Paul Smith
- Correction of incorrect submission in v0.1.4 of the fix to the
check.inputs = FALSE example
- Fixed simmilar ASAN and valgrind error in unit tests
- Stopped RUnit tests writing files which caused CRAN build errors due
to directory permissions
2018-07-01 Paul Smith
- section Maintainer updated
- Fixed check.inputs = FALSE example which was passing incorrect
format variables resulting in ASAN and valgrind errors
- Moved function documentaion to use roxygen2 for generation
2014-01-03 Aleksandar Spasojevic
- sspir::kfilter reference removed because package sspir has been
- section Maintaner updated since ZHAW - Institute of Data Analysis
and Process Design is maintaing FKF now
2012-03-20 David Luethi
- RUnit added as dependancy
2010-10-17 David Luethi
- Error in plot.fkf: Standardized residuals were calculated
- plot-method of fkf-objects with type == ‘resid.qq’ puts the number
of the variable in the title (main = …)
- fkf.Rd, plot.fkf.Rd: Documentation extended, examples added.
- R unit test added.
2010-12-19 Philipp Erb
- Fixed minor inconsistency in input check messages of fkf()
2009-09-28 Philipp Erb
- NA-values in the vector of response variables yt are now supported.
In the filter recursions the observation vector yt with missing values
will be reduced to the vector yt- without missing values and the
measurement equation will be adjusted accordingly.
2009-07-30 David Luethi
- c-function ‘cfkf’: Warning for non-finite ‘logDet’ or ‘mahalanobis’
removed (too verbose as information is contained in the output ‘status’
and also documented). Potential warnings occurring in the Cholesky
factorization are more informative now.
2009-06-26 David Luethi
- Minor corrections in the documentation.
- Function ‘fkf’ scans for missing arguments now.
2009-02-10 David Luethi