Implementation of "approximate" or "random" permutation tests with the argument permutation.test in the functions change.point, foster.test, and N.test. The tests use samples with repetition of permutations, they do not use all possible permutations. Unlike previous exact, asymptotic, or Monte Carlo tests, permutation tests still ensure test size by introducing a matrix X with dependent columns as an argument.
Function records when plot = FALSE returned the wrong backward record values. It returned the values of the forward series on the backward record times. This issue was fixed.
Function L.record returned a matrix instead of a list when all series had the same number of records. Function R.record returned incorrect output when all series had the same number of records or there were only one series. These issues were fixed and both functions return a list in any case.
Function series_untie now admits NA or NaN values without returning NaN's.
Changes in Version 2.1.2 (15 March 2023)
MODIFICATIONS
‘CITATION’ and references update for the new JSS publication of Castillo-Mateo, Cebrian and Asin (2023) doi:10.18637/jss.v106.i05.
Changes in Version 2.1.1 (03 November 2022)
NEW FEATURES
New dataset TX_Zaragoza29F that includes Zaragoza series with the observations from the 29th of February.
Function series_uncor now admits any test introduced as an argument to test correlation or dependence across columns. Default is cor.test, which is the test used in previous versions.
Confidence intervals (CIs) are renamed to reference intervals (RIs) in the ‘MANUAL’ and legends for being a more appropriate definition for these intervals (no argument is modified).
BUG FIXES
Functions N.plot and p.plot displayed all four types of records in the legend, even when only one type of record was plotted. Now the legend only shows the type of records that are plotted.
Changes in Version 2.1.0 (05 August 2021)
NEW FEATURES
New function change.point to perform the Castillo-Mateo change-point detection tests was added.
Previously, if the series had NA values, all the record indicators from that time took NA value. This made the functions return NA p-values, etc. Now the NA values are treated as non-records, i.e., I_t = 0, unless they are the first observation, i.e., I_1 = 1.
records function only allowed to show points, now it also shows lines.
MODIFICATIONS
The three missing values of the TX_Zaragoza dataset now take the value NA, in previous Versions these values were imputed. This change does not affect the ZaragozaSeries dataset.
The ‘MANUAL’ help pages were updated.
The ‘VIGNETTES’ file was updated.
Changes in Version 2.0.0 (13 February 2021)
NEW FEATURES
New function brown.method to perform the the Brown's method in the (weighted)-number-of-records was added.
New function fisher.method to perform Fisher's method was added.
New function R.record to compute record values.
New function rcrm to generate random samples of the classical record model.
New function series_ties to summarize the presence of record ties (or weak records) mainly due to rounding effects.
New dataset Olympic_records_200m with record times and record values of the Olympic 200-meter race.
New function series_record to apply the tools within RecordTest if only the record times are available and not the complete series.
New argument weak if weak records are also computed in functions that compute the record random variables: I.record, L.record, N.record, etc.
MODIFICATIONS
The ‘MANUAL’ help pages were updated.
The ‘VIGNETTES’ file was updated.
All functions have undergone a change in their arguments, providing new features in most cases.
The functions foster.test and foster.plot now admit more statistics based on different combinations of records.
The function series_uncor has change the criteria of considering the first and last columns uncorrelated, thus slightly different results can come from previous Versions. As new features it admits more types of uncorrelation and more arguments.
The dataset TX_Zaragoza has been updated with more data.
The matrix ZaragozaSeries has more rows and one more column according to the changes in TX_Zaragoza and series_uncor.
Functions have been joined or renamed as follows:
From L_global.test renamed to global.test, and now a global statistic for different types of record can be computed for the tests with two-sided alternatives p.chisq.test, p.regression.test, lr.test and score.test.
From N_normal.test, N_exactPB.test and P_exactPB.test joint to N.test, and new features added.
From P_regression.test and P_regression.plot renamed to p.regression.test and p.plot, and new features added.
From P_chisq.test renamed to p.chisq.test.
From L_lr.test and L_lm.test renamed to lr.test and score.test.
From M.record renamed to S.record.
BUG FIXES
The p-values in Monte-Carlo and Poisson-Binomial tests were computed with the expression P(X > X_0) instead of P(X \ge X_0) in all previous Versions, this problem was fixed.
Function foster.plot showed the evolution of the D statistic improperly, this problem was fixed.
Changes in Version 1.0.1 (05 October 2020)
NEW FEATURES
The ‘README’ and ‘VIGNETTES’ files were added.
New function foster.test to perform the Foster-Stuart and Diersen-Trenkler tests was added.
New function foster.plot to perform plots based on the Foster-Stuart and Diersen-Trenkler statistics was added.
New function L_global.test to perform likelihood ratio and Lagrange multiplier tests on record occurrence was added.
New function L_lm.test to perform Lagrange multiplier test on record occurrence was added.
New function series_untie to deal with record ties was added.
Support for parallel computing was added in Monte Carlo functions: L_global.test, L_lm.test, L_lr.test.