Statistical Tolerance Intervals and Regions

Statistical tolerance limits provide the limits between which we can expect to find a specified proportion of a sampled population with a given level of confidence. This package provides functions for estimating tolerance limits (intervals) for various univariate distributions (binomial, Cauchy, discrete Pareto, exponential, two-parameter exponential, extreme value, hypergeometric, Laplace, logistic, negative binomial, negative hypergeometric, normal, Pareto, Poisson-Lindley, Poisson, uniform, and Zipf-Mandelbrot), Bayesian normal tolerance limits, multivariate normal tolerance regions, nonparametric tolerance intervals, tolerance bands for regression settings (linear regression, nonlinear regression, nonparametric regression, and multivariate regression), and analysis of variance tolerance intervals. Visualizations are also available for most of these settings.


News

####################################### 2017-02-03 (Version 1.3.0) ####################################### *Updated "Title" and "Description" to reflect new CRAN policies. *Added an error message if a vector of length less than 2 is used for the data argument in nptol.int(). Thanks to Maxim Nazarov for the suggestion. *Fixed bug in dnhyper() for case where x=0. Thanks to Jerry Lewis for reporting this bug. *Added functions for constructing simultaneous tolerance limits for normal populations. This includes a k-factor function [K.factor.sim()] and a function to construct the simultaneous normal tolerance limits [simnormtol.int()]. Thanks to Andrew Landgraf for providing his code for the K.factor.sim() function. *Updated some references.

####################################### 2016-02-17 (Version 1.2.0) ####################################### *Added functions for the discrete Pareto distribution. This includes distribution functions [ddpareto(), pdpareto(), qdpareto(), and rdpareto()], a maximum likelihood function [dpareto.ll()], and a function for calculating approximate tolerance intervals [dparetotol.int()]. Thanks to Mehran Naghizadeh Qomi and Azadeh Kiapour for initial versions of these functions and for thoroughly testing the final functions. *Fixed issue with loading stats4 package for the zm.ll() and poislind.ll() functions. *Added norm.OC(), a function for producing OC-type curves to illustrate how values of the k-factors for normal tolerance intervals, confidence levels, and content levels change as a function of the sample size. *Corrected error in anovatol.int() function. The k-factor calculation in the previous version used an incorrect value for the degrees of freedom in the formula. *Fixed typo in norm.ss() documentation. *Modified K.table() code to allow user to specify any of the ways to compute the underlying normal k-factors. *Modified regolt.int() code to allow predictions when weighting is used. *Fixed typos in dpoislind() and ppoislind() functions regarding the logarithm options.

####################################### 2015-12-11 (Version 1.1.1) ####################################### *Added bayesnormtol.int(), a function to construct Bayesian normal tolerance intervals. *Added norm.ss(), a function with various strategies for sample size determination using normal tolerance intervals. *Updated some references.

####################################### 2015-04-15 (Version 1.1.0) ####################################### *Added functions for the discrete Poisson-Lindley distribution. This includes distribution functions [dpoislind(), ppoislind(), qpoislind(), and rpoislind()], a maximum likelihood function [poislind.ll()], and a function for calculating approximate tolerance intervals [poislindtol.int()]. *In the K.factor() function, the upper limit in the integrate function for the "EXACT" method was changed from 10 to 5. A numerical error would sometimes occur for extreme values of alpha, which appears to be fixed by this change. If any errors pertaining to the integrate function are obtained because of this change, please contact me and I will try to dig into this further.

####################################### 2014-09-03 (Version 1.0.0) ####################################### *Fixed small bug in the qnhyper() function. Fixed an incorrect assignment in the dnyhper() and the pnhyper() functions. *Added the "EX" option for an exact method calculation in the neghypertol.int() function. *Added the "EX" option for an exact method calculation and fixed how lower and upper bounds were defined in the hypertol.int() function. *Added the "YM" method to nptol.int(), which is the Young-Mathew method for improved nonparametric tolerance intervals based on interpolated and extrapolated order statistics. *Updated some documentation. *Updated my contact e-mail as maintainer.

####################################### 2013-10-11 (Version 0.5.3) ####################################### *Updated the "OCT" option in the K.factor() function to allow the user to use the degrees of freedom option, "f". *Fixed a bug in the bonftol.int() function. The "alpha" argument was not passed correctly to the type of tolerance interval being calculated. Thanks to Alson Look for reporting this bug. *Updated references.

####################################### 2012-09-23 (Version 0.5.2) ####################################### *The mvtol.region() function was updated to include many different ways to approximate the tolerance factor for multivariate normal data. These were mainly included for comparative purposes. The default method ("KM") is the Krishnamoorthy-Mondal method, which was the method used in all previous versions of the 'tolerance' package. *Fixed a bug in the K.table() function. When summarizing by "n", the wrong indexing was used, thus yielding incorrect results.

####################################### 2012-08-28 (Version 0.5.1) ####################################### *The functions related to Zipf-Mandelbrot distributions have been updated. The distribution functions (i.e., qzipfman, pzipfman, dzipfman, and rzipfman) have all been updated to be more efficient. Moreover, there were some computational issues for extreme quantiles in these functions. Namely, negative density values would be returned. There have been corrections for these bugs, plus I have included an interpolation procedure when calculating very large quantities for the zeta distribution. Deleted the "exact" option in the respective functions. *The zipftol.int function has been updated to include the changes in the distribution functions. There is also a new argument that allows the user to specify how to handle ties. Namely, if the limit is calculated at a category that has the same frequency as other adjacent categories, then you can either take the most extreme of those tied categories (which would be the lowest of the tied categories for the lower limit and the highest of the tied categories for the upper limit) or simply take the category that was estimated by the procedure. This latter approach would only be used if there is some scientific knowledge that could basically guide the user with setting an ordinal ranking of the tied categories.

####################################### 2012-07-19 (Version 0.5.0) ####################################### *Added code for using a fiducial approach to estimate tolerance intervals for the function of two binomial proportions (fidbintol.int), two Poisson rates (fidpoistol.int), and two negative binomial proportions (fidnegbintol.int). *Added three new options for estimating the two-sided K-factor for the normal setting: one due to Howe, one due to Krishnamoorthy and Mathew, and one for controlling the tail probabilities.
The exact method has also been made more efficient. Also, corrected a small error in the normtol.int function that did not allow the user to call all of the possible methods as cited in the documentation. The K.factor() function was also vectorized to accommodate a vector of sample sizes. *Added a function that calculates Appell's hypergeometric function of the first kind (F1), distribution functions for the difference between two proportions (qdiffprop, pdiffprop, ddiffprop, and rdiffprop), and code for estimating tolerance intervals for the difference between two proportions using a fiducial-based approach (diffproptol.int). *Condensed the Zipf, Zipf-Mandelbrot, and zeta distribution functions into the qzipfman, pzipfman, dzipfman, and rzipfman functions. This simultaneously corrected a minor bug in the original distribution functions of the three distributions. Searches on any of these distributions will now point to these new functions. Future work will likely be done on these functions to make them more efficient. *Added distribution functions (i.e., density, cumulative distribution, quantile, and random generation) for the negative hypergeometric distribution. *Added the Krishnamoorthy-Mathew approach for computing the upper tolerance bound for the 2-parameter exponential distribution. *Added neghypertol.int() function for calculating tolerance limits for negative hypergeometric random variables. *Added hypertol.int() function for calculating tolerance limits for hypergeometric random variables. *Added new options for how the confidence intervals for the binomial proportions are calculated in the bintol.int function. Specifically, the methods "PR", "PO", "CL", "CC", and "CWS" have been added, which are for the probit transformation, Poisson parameterization, complementary log transformation, continuity corrected large sample approach, and continuity corrected Wilson's approach, respectively. *Added new options for how the confidence intervals for the Poisson rates are calculated in the poistol.int function. Specifically, the methods "CC", "VS", "RVS", "FT", and "CSC" have been added, which are for the continuity corrected large sample approach, the variance- stabilization approach, the recentered variance stabilization approach, the Freeman-Tukey method, and the continuity corrected score method, respectively. *Completely overhauled the negbintol.int function, which was not providing good estimates. The new version provides many ways to estimate the negative binomial proportion confidence intervals, similar to how the binomial and Poisson tolerance limits are constructed. An article has also been submitted outlining the procedure. *The use of the ppois function was misspecified for the lower tolerance limit in the poistol.int function. This resulted in the reported lower limit being larger by 1 value. This has been corrected. *Fixed the umatol.int function to allow the case of x = 0. *Fixed a small typo in the negbintol.int documentation. *Put maximum tolerance limits for the discrete distributions at Inf instead of 0.

####################################### 2011-12-07 (Version 0.4.0) ####################################### *Added negbintol.int() function for calculating tolerance limits for negative binomial random variables. *Updated poistol.int() to also include the score method as a way to estimate the tolerance limits. *Added distribution functions (i.e., density, cumulative distribution, quantile, and random generation) for zeta, Zipf, and Zipf-Mandelbrot distributions. *Added zm.ll() function for estimation of the shape parameter(s) in the zeta, Zipf, and Zipf-Mandelbrot distributions. *Added zipftol.int() function for calculating tolerance intervals for the zeta, Zipf, and Zipf-Mandelbrot distributions. *Corrected bug in p2exp() and q2exp() functions. *Corrected how the output is displayed for the nptol.int() function when method="HM". *Fixed rounding error in the acc.samp() function and added clarifying text in the acc.samp() documentation. *Fixed estimation issue in exttol.int() function. The Newton-Raphson algorithm had an error when parameter values were relatively large.

####################################### 2011-04-03 (Version 0.3.0) ####################################### *Changed the output for regtol.int(), nlregtol.int(), and npregtol.int() from class "matrix" to "data.frame". *Updated K.factor() to also include the option method="EXACT", which does an exact calculation of two-sided k-factors by numerically solving the necessary integral calculation. *Reflected the new exact method for calculating the k-factors above in the gamtol.int() and anovatol.int() functions. *The function bonftol.int() was added to approximate two-sided tolerance intervals that control the proportion of the population in the tails. *The function diffnormtol.int() was added for calculating one-sided tolerance limits for the difference between two independent normal random variables. An exact calculation is performed when the variance ratio is known, while various estimation methods are used when the variance ratio is unknown. *Updated references.

####################################### 2010-10-07 (Version 0.2.3) ####################################### *Corrected acc.samp() function so that the output is of class "matrix". *Changed the exp2tol.int() function so that it does not automatically truncate the lower tolerance limit at 0. *The function paretotol.int() was added to estimate tolerance intervals for the Pareto distribution as well as the power distribution.

####################################### 2010-07-30 (Version 0.2.2) ####################################### *Updated references. *Fixed minor typos in documentation.

####################################### 2010-05-16 (Version 0.2.1) ####################################### *Started a "NEWS" file for changes in the package. *Changed the way the k-factor for the two-sided linear and nonlinear regression tolerance intervals are calculated. Before, an approximation was used. A more accurate method (presented at the bottom of p. 70 of Krishnamoorthy and Mathew (2009)) is used. *Fixed a serious bug with the way nlregtol.int() checks that the pseudo-design matrix is, in fact, invertible. *The function mvregtol.region() was added to estimate multivariate multiple linear regression tolerance factors.

####################################### 2010-05-02 (Version 0.2.0) ####################################### *New function anovatol.int() performs tolerance interval calculations for balanced ANOVA. *New function np.order() performs sample size determination for tolerance limits based on order statistics. *New function umatol.int() performs uniformly most accurate upper tolerance limits for the Poisson, binomial, and negative binomial distributions. *Updated K.factor() to also include the option method="ELL" for the Ellison correction method. *Updated acc.samp() to include the option RQL. Documentation has also been updated to provide more detailed explanations of the function's arguments. *Updated exttol.int() to also perform calculations for the Gumbel distribution for the maximum. Before, this function only did calculations for the Weibull distribution and the Gumbel distribution for the minimum. *The portion of code for gammatol.int() when side=2 was incorrect. It has now been corrected. *laptol.int(), logistol.int(), uniftol.int(), and exttol.int() all now include the option side=2 to estimate two-sided tolerance intervals. The option simply calculates a Bonferroni approximation for the two-sided setting. *uniftol.int() has now been written to for the general uniform case and not simply for the setting where the distribution has a lower bound of 0. *plottol() has been updated so that pch=19 is used by default for all of the scatterplots. It also includes functionality for a plot pertaining to output from the new anovatol.int() function. *Fixed minor typos in the documentation (e.g., the Poisson tolerance intervals documentation originally said ``Poison"). *Added some additional references to some of the documentation.

####################################### 2009-06-29 (Version 0.1.0) ####################################### *The 'tolerance' package is officially launched.

Reference manual

It appears you don't have a PDF plugin for this browser. You can click here to download the reference manual.

install.packages("tolerance")

2.0.0 by Derek S. Young, a year ago


Browse source code at https://github.com/cran/tolerance


Authors: Derek S. Young [aut, cre]


Documentation:   PDF Manual  


Task views: Probability Distributions


GPL (>= 2) license


Imports MASS, rgl, stats4


Imported by adamethods, fdasrvf, moezipfR, zipfextR.

Suggested by PredictionR.


See at CRAN