Sign up to receive email updates about extRemes. These updates are highly infrequent, and typically minor updates such as for version 1.55 are not announced, so check here if you find that your version of extRemes does not work with your version of R. You will also be able to unsubscribe from this list using the same link.
Changes in extRemes with version 1.57
Further fixes to return.level including a switch to using the exact gradient for the return level function when computing the delta method confidence intervals. Previously, the difference approximation was used.
No asymptotic results are used for finding return levels for return periods of less than one year. Only return periods greater than 1 year can be computed with this function.
Note that this function is not called by any of the GUI windows, which continue to use ismev routines.
Changes in extRemes with version 1.56
Fixed some problems with the command-line function (not used with the GUIs) 'return.level' designed to allow the user to pass particular desired return periods, and receive the return levels (and delta method confidence intervals, if desired) as output instead of simply plotting them.
Changes in extRemes with version 1.55
Only one minor change here to (with any luck) remove a warning found with 'R CMD check' of the pre-release of R 2.5.0.
Changes in extRemes with version 1.54
No noticeable changes have taken effect here. Just some minor modifications to keep up with newer versions of R. Note that the package may no longer work with older versions of R. It will be necessary to update to the latest version of R for the package to work.
Changes in extRemes with version 1.53
Mostly cosmetic and probably will not be noticed by the user, but I did fix a bug whereby when one transforms data to log daily returns, a missing value was left in the final spot, which is subsequently not handled by numerous other functions. So, any missing values (either at the end, or created from the transformation) are replaced by the mean of the resulting returns.
Changes in extRemes with version 1.52
One can now obtain return levels in an R object by typing a new object name in the Save As (optional) field of the Return Level Plot GUI (i.e., Click Plot followed by Return Level). A list object is returned, and assigned to an object with the Save As name, with values of the return period (period), the return values (values), and the lower and upper confidence intervals (if any).
When obtaining confidence intervals via the profile likelihood method, the window now remains open with the values last entered remaining. This saves time when trying various different limits, etc. It is helpful for finding confidence intervals for multiple return levels.
Changes in extRemes with version 1.51
The point-process algorithm has been fixed so that inclusion of covariates works correctly, if slowly. The function fpp has been added to replace use of the ismev function pp.fit, which uses an incorrect approximation to the exceedance rate when including covariates.
Added some code written by Chris Ferro. Specifically, the following functions have been added: boot.matrix, boot.sequence, decluster.runs, decluster.intervals, and exi.intervals: and each function has an associated help file. In particular, the following features are now included with extRemes.
Intervals estimator for extremal index now gives (bootstrap) confidence limits.
It is now possible to decluster using intervals declustering (but not directly via GUIs).
The runs declustering algorithm is now fast.
Run length estimator has been fixed.
A few typos in dataset help files have been corrected.
Changes in extRemes with version 1.50
The primary changes in the new version (1.50) of extRemes are:
Pertinent code executed by the GUIs is written to a log file (extRemes.log) ...
more on this
All other messages/output are displayed on the main R console instead of the console
and/or the main toolkit dialog window.
The current R workspace can now be saved from the main toolkit dialog ...
more on this
When data are read into R via the toolkit dialog, the current R workspace is saved, and a
statistical summary of the data is given (e.g., upper and lower quartiles, min, max, mean, etc.).
The class "ev.data" is now called "extRemesDataObject".
There is a new function called as.extRemesDataObject. Use the command
help( as.extRemesDataObject) for more on this function.
Much more information is given when help( extRemes) is invoked.
There is no longer a time series option under Plot ... more on this
MLE estimation for routines that require the R function optim now defaults to the "BFGS"
optimization method if no method is selected (instead of returning an error).
Pertinent code executed by the GUIs is written to a log file (extRemes.log), enabling the user
to view the actual R commands carried out by each window action. The file is written to the directory
whence R was launched. In Windows, the location of this directory may not be obvious (e.g., if a
shortcut is used to open the file). Use the command getwd() from the R console prompt in order
to find the path to this directory. Additionally for Windows, the extension '.log' may be invisible; look
for the text file 'extRemes'.
Note: The log file can get to be very large as new commands are appended to the
file. It is possible to clear this log file from the GUI windows at any time--the file will still be there,
but will only display a messaging giving the date and time it was cleared. To do this, click on
File > Clear log file from the main toolkit dialog.
Furthermore, note that list components are accessed using the full list.name[["component.name"]] notation
because the shortcut notation, list.name$component.name, does not work properly when using eval;
which is necessary in order to send the (exact) commands to a file. For example, to access the "data" component
from a list, say it is called "Fort", one can do either of the following commands.
> Fort[["data"]] # As would be seen in extRemes.log > Fort$data # As can be done from the command line.
More help files are included with extRemes. In particular, there are help files for every dataset
provided with the toolkit. Use the command help( extRemes) to see a list of all of these
datasets.
There is no longer a time series option under Plot. It was never a true time series. In particular,
it did not make use of the R package ts. It had been a scatter plot with lines connecting points; though
it was possible to plot a time series in this way. Instead, there is now a line option on the scatter plot
dialog, which provides the same functionality as before.
