The issue described at https://github.com/RcppCore/Rcpp/issues/1287 has been fixed to avoid WARNINGs from CRAN checks on some platforms. Thank you to Dirk Eddelbuettel for providing the fix so quickly!

Fixed issues with the incorrect use of in some Rd files.

If the argument

`k = 0`

is supplied to`kgaps()`

then an estimate of 1 is returned for the extremal index for any input data. For this very special case the estimated standard error associated with this estimate is set to zero and confidence intervals have a width of zero.Corrected a typing error in the description of

`uprob`

in the documentation for`plot.choose_uk()`

and`plot.choose_ud()`

.The unnecessary C++11 specification has been dropped to avoid a CRAN Package Check NOTE.

README.md: Used app.codecov.io as base for codecov link.

Create the help file for the package correctly, with alias exdex-package.

- A new estimator has been implemented, based on what we will call the D-gaps model of Holesovsky, J. and Fusek, M. Estimation of the extremal index using censored distributions. Extremes 23, 197–213 (2020). doi: 10.1007/s10687-020-00374-3

The value returned by

`nobs.kgaps()`

was incorrect in cases where there are censored K-gaps that are equal to zero. These K-gaps should not contribute to the number of observations. This has been corrected.In cases where the data used in

`kgaps`

are split into separate sequences, the threshold exceedance probability is estimated using all the data rather than locally within each sequence.A

`logLik`

method for objects inheriting from class`"kgaps"`

has been added.In the (unexported, internal) function

`kgaps_conf_int()`

the limits of the confidence intervals for the extremal index based on the K-gaps model are constrained manually to (0, 1) to avoid problems in calculating likelihood-based confidence intervals in cases where the the log-likelihood is greater than the interval cutoff when theta = 1.In the documentation of the argument

`k`

to`kgaps()`

it is noted that in practice`k`

should be no smaller than 1.The function

`kgaps()`

also return standard errors based on the expected information.In the package manual related functions have been arranged in sections for easier reading.

Activated 3rd edition of the

`testthat`

package

- The functions
`kgaps()`

,`kgaps_imt()`

and`choose_uk()`

can now accept a`data`

argument that- is a matrix of independent subsets of data, such as monthly or seasonal time series from different years,
- contains missing values, that is,
`NA`

s.

- A new dataset
`cheeseboro`

is included, which is a matrix containing some missing values. - In addition to
`kgaps()`

, the functions`kgaps_imt()`

and`choose_uk()`

now have an extra argument`inc_cens`

, which allows contributions from censored K-gaps to be included in the log-likelihood for the extremal index. - The default value of
`inc_cens`

in`kgaps()`

(and in`kgaps_imt()`

and`choose_uk()`

) is now`inc_cens = TRUE`

.

- Plot and print methods have been added for objects of class
`"confint_gaps"`

returned from`confint.kgaps()`

. - In
`confint.spm()`

and`confint.kgaps()`

the input confidence`level`

is included in the output object.

- An overloading ambiguity has been corrected to ensure installation on Solaris.