Detection of Structural Changes in Climate and Environment Time
Series
Tools for automatic model selection and diagnostics for Climate and Environmental data. In particular the envcpt() function does automatic model selection between a variety of trend, changepoint and autocorrelation models. The envcpt() function should be your first port of call.
News
Version 1.1.1
Changes to previous version
Fixed bug in envcpt function that meant that if a non-changepoint model was not specified in the list of models to fit, and no changepoints were detected in the changepoint model then the log-likelihood was recorded in the output as NA instead of the value fitted.
Version 1.1
Changes to previous version
Added option to envcpt to choose a subset of models to run.
The subset selection also follows through to plotting whereby only the models fitted are plotted.
BIC method added.
AICweights function added to calculate the AIC weights for the models fitted.
Removed dependency on forecast package in favour of directly using arima function from stats package. (slimming reverse dependencies)
Added new tests for the new functionality
Added EnvCpt paper to references
Version 1.0
Changes to previous version
Coding for cpt.reg has been implemented in C.
Changepoints in normal linear regression takes advantage of the Fortran dqrls command that is called with lm.fit. As such, potential multicollinearity is appropriately managed.
cpt.reg is available for any number of regressors.
The file LMregression.R is no longer used and so is removed.
Version 0.1.1
Changes to previous version
Did corrections for CRAN submission.
Corrected plot(.,type='aic') to do the correct axis drawing when yaxt is specified.