Changelog • dailymet

New in version 0.1.7

In quantMax the misleading warning in quantMax is no longer thrown and the deailed steps of modelMatrices are no longer shown when trace = 0.
In the output of quantMax, the confidence level is added as a column Level, for compatibility with the quantMax method of the class TVGEV" of the NSGEV package.

The arguments newdata and prob of the quantMax method for the class "pgpTList" were not used.

In both in rqTList and pgpTList it is now possible to use quite arbitrary “design” functions to create variables entering in the formula(s) defining the model objects. The designVars function has been added to simplify the managment of the variables in fitting, prediction and simulation steps.
A new generic quantMax and the corresponding method for the "pgpTList" class have been added. This method should be used in place of quantile to avoid possible confusions between the quantile of the maximum and the marginal quantiles.

The quantile method for the class "pgpTList" now computes the confidence intevals by the “delta method”. The quantiles and confidence intervals are computed for a quite large number of probabilities in order to build nice-looking plots.
For the objects with class "quantile.pgpTList" as created when applying the quantile method, there now exist an autoplot method and a format method. The first builds a ggplot, the second makes a table with suitably rounded quantiles and selecting round probabilities as is often needed in reports.
The Intro vignette has been enhanced to illustrate the use of the new methods. The secion Investigating the distribution of the maximum should be instructive for the users.

In the predict.pgpTList there was an error in the determination of the maximum of the thresholds. This could have an impact on the result of the quantile method although it seems very small in practice.

The simulate.pgpTList can now be used to compute the quantiles of the maximum M over the prediction/simulation period.

The predict method for the class "PgpTList" provides new variables on output.
The new method quantile for the classes "PgpTList" and "predict.PgpTList" allows to investigate the tail distribution of the maximum on a “new” period of interest.

The predict method for the class "PgpTList" now accepts a newdata argument which can be simply a vector with class "Date".
New method simulate for the class "PgpTList". It has a newdata argument similar to that of the predict method.
The predict and the simulate methods return objects having the S3 classes "predict.pgpTList" and "simulate.pgpTList" for which some methds are available, such as autoplot.
The package no longer depends on NSGEV and potomax from which it imported the GEV and the GPD2 distributions respectively. These distributions are now imported from the nieve package.