This is the version of the package under development. To use the most recent stable version from CRAN, run
install.packages("funtimes")
library(funtimes)
To get this version from GitHub, run
devtools::install_github("vlyubchich/funtimes")
library(funtimes)
- Remove defunct functions.
- Describe the statistics in causality_pred()
- Added new function
r_crit()
to compute critical values of correlations.
- Replace the
BIC
argument in AR estimation withic
, to easily switch to AIC. The directly affected functions arenotrend_test()
,sync_test()
, andwavk_test()
. - The functions
i.tails()
andq.tails()
are marked defunct and renamed astails_i
andtails_q
. - Remove dependence on the package
Jmisc
. - The function
ccf_boot()
now has the parallel option, thanks to suggestions from Erika Koontz and Benjamin Malmgren. The function also switched to using normal bootstrap intervals to improve their symmetry. Optional smoothing was added for extra smoothness, which prompted to omit$p$ -value calculation. - In the function
causality_pred()
, remove the fast bootstrap of paired out-of-sample errors as under-performing; make sure the AR-order selection is done on the training set; try extra statistics; correct the restricted model for bootstrapping (include intercept).
- Added functions for out-of-sample Granger causality testing (based on prediction errors), with an option to restrict (disregard) near-contemporaneous lags:
causality_pred()
andcausality_predVAR()
.
- Removed dependency on the package
TDA
by omitting the TDA-based clustering functionTopoCBN()
. - Added a vignette on clusters of time series.
mcusum_test
now runs faster, has the option to use heuristic shortened bootstrapping, and an option to setm <= length(k)
(used to bem = length(k)
) thanks to the contributions from Alexander Brenning.
- Deprecated functions (with dots) marked as defunct.
- New functions:
AuePolyReg_test()
,ccf_boot()
,cumsumCPA_test()
,GombayCPA_test()
, andTopoCBN()
. - Fixed a bug in
beales()
thanks to an email from Dave Lorenz.
- Use underscores in function names instead of dots. Mark those functions as deprecated.
- Add vignettes for Beale's estimator and trend tests.
- Add a package overview section in the documentation.
- Format R code according to R style diagnostics (such as whitespaces).
- Improve computational efficiency of
purity()
function, inspired by Brian Simmons's suggestions from 2019-06-12 about handling zero cases when all matches have been found.
See GitHub commits before 2019-04-15