Simulation and Resampling Methods for Epistemic Fuzzy Data.

FuzzySimRes

The goal of FuzzySimRes, a library written in R, is to simulate synthetic fuzzy data and provide so-called epistemic bootstrap procedures.

The simulation procedures are useful to generate synthetic samples that can be applied in statistical inference (see, e.g., (P. Grzegorzewski, Hryniewicz, and Romaniuk 2020a, 2020b; Parchami, Grzegorzewski, and Romaniuk 2024)) and consist of fuzzy numbers (e.g., triangular or trapezoidal ones).

The epistemic bootstrap is used for epistemic fuzzy data (see (Couso and Dubois 2014)) and it allows (see (P. Grzegorzewski and Romaniuk 2021; Przemyslaw Grzegorzewski and Romaniuk 2022)):

• application of special resampling procedures,
• application of real-valued statistical tests for fuzzy data,
• estimation of statistical characteristics (like the mean) of a fuzzy sample.

Additionally, the special epistemic fuzzy real-life data set is provided (see (Faraz and Shapiro 2010)) which consists of triangular fuzzy numbers.

The fuzzy data used in this package should be defined as the variables introduced in (Gagolewski and Caha 2021) (as, e.g., triangular fuzzy numbers).

The following procedures are available in the library:

• Epistemic bootstrap procedures:
  • EpistemicBootstrap - the standard epistemic bootstrap,
  • AntitheticBootstrap - the antithetic epistemic bootstrap.
• Random generation of the initial samples:
  • SimulateFuzzyNumber - generation of a single fuzzy number (triangular, trapezoidal, PLFN one),
  • SimulateSample - generation of the whole fuzzy sample using various random distributions.
• Applications of the epistemic bootstrap:
  • EpistemicEstimator - estimation of the given statistical characteristic together with its SE/MSE using the fuzzy sample and epistemic bootstrap,
  • EpistemicMean - estimation of the mean using the fuzzy sample and epistemic bootstrap,
  • EpistemicCorrectedVariance - estimation of the variance with the special corrected formula using the fuzzy sample and epistemic bootstrap,
  • EpistemicTest - estimation of the p-value of the selected real-valued statistical test using the fuzzy sample and epistemic bootstrap,
  • AverageStatisticEpistemicTest - estimation of the p-value of the selected real-valued statistical test using the fuzzy sample, epistemic bootstrap, and averaging statistics method,
  • MultiStatisticEpistemicTest - estimation of the p-value of the selected real-valued statistical test using the fuzzy sample, epistemic bootstrap, and multi-statistics method,
  • ResamplingStatisticEpistemicTest - estimation of the p-value of the selected real-valued statistical test using the fuzzy sample, epistemic bootstrap, and resampling statistics method.
• Additional procedures:
  • CombinePValues - combining the p-values of a statistical test into a single value,
  • KSTestCriticalValue - calculation of the p-value for a given critical value in the case of the one- and two-sample Kolmogorov-Smirnow test.
• Dataset:
• controlChartData - the real-life fuzzy epistemic data concerning electronic circuit thickness.

Installation

You can install the latest development version of FuzzyResampling with:

library(devtools)
install_github("mroman-ibs/FuzzySimRes")

You can install the latest stable version from CRAN with:

install.packages("FuzzySimRes")

Examples

# set seed

set.seed(12345)

# load library

library(FuzzySimRes)


# generate the initial sample consisting of 5 trapezoidal fuzzy numbers

sample1 <- SimulateSample(5,originalPD="rnorm",parOriginalPD=list(mean=0,sd=1),
incrCorePD="rexp", parIncrCorePD=list(rate=2),
suppLeftPD="runif",parSuppLeftPD=list(min=0,max=0.6),
suppRightPD="runif", parSuppRightPD=list(min=0,max=0.6),
type="trapezoidal")

# apply the epistemic bootstrap with 10 cuts

EpistemicBootstrap (sample1$value, cutsNumber=10)
#>               X1       X2         X3         X4          X5
#> 0.3616 1.1497756 1.892964 -0.4412416 -0.4123010 -0.21887696
#> 0.8688 1.0789047 2.256274  0.8043010 -0.2897991  0.21447852
#> 0.9042 1.0821638 1.558882 -0.5234141 -0.4201384  0.01919864
#> 0.6174 0.9770422 1.871995  0.2864493 -0.4820862  0.22506592
#> 0.134  1.4685182 1.471085  1.1883406 -0.7735033 -0.53509789
#> 0.7822 1.6920152 2.316149  0.3981874 -0.3132989  0.29258249
#> 0.4292 1.6567735 1.597893  0.1417311 -0.4820543  0.24728641
#> 0.9273 1.2565657 2.537325 -0.1589284 -0.6728166 -0.22298375
#> 0.7732 1.1521309 2.275875 -0.2135284 -0.7091520  0.07312683
#> 0.2597 1.5825929 2.525235 -0.8153841 -0.7529021  0.17019910

# apply the antithetic bootstrap with 10 cuts

AntitheticBootstrap (sample1$value, cutsNumber=10)
#>              X1       X2          X3         X4          X5
#> 0.5138 1.715099 2.348629  0.84791420 -0.4155065 -0.10278774
#> 0.7201 1.657023 2.315396  0.75226172 -0.2950119 -0.18807131
#> 0.7499 1.525241 1.529150  0.04191831 -0.4140887  0.28010819
#> 0.0956 1.036562 1.548077  0.88263662 -0.2954918 -0.28473900
#> 0.3978 1.123620 2.289173  0.08373510 -0.4331157 -0.19568336
#> 0.2945 1.423709 1.622644  0.68312991 -0.4069346  0.13470515
#> 0.6173 1.300673 2.154367  0.27146818 -0.6049461  0.30957665
#> 0.9743 1.153848 1.880561  0.76471178 -0.5728187 -0.29215072
#> 0.6182 1.384238 2.107480  0.30352891 -0.6221601 -0.08907579
#> 0.5214 1.574392 1.977895 -0.28340154 -0.3810230  0.25160955

# calculate the mean using the standard epistemic bootstrap approach

EpistemicMean(sample1$value,cutsNumber = 30)
#> $value
#> [1] 0.60163
#> 
#> $SE
#> [1] 0.1570901
#> 
#> $MSE
#> [1] NA

# calculate the median using the antithetic epistemic bootstrap approach

EpistemicEstimator(sample1$value,estimator="median",cutsNumber = 30,bootstrapMethod="anti")
#> $value
#> [1] 0.136565
#> 
#> $SE
#> [1] 0.2187421
#> 
#> $MSE
#> [1] NA


# generate two independent initial fuzzy samples

list1<-SimulateSample(20,originalPD="rnorm",parOriginalPD=list(mean=0,sd=1),
incrCorePD="rexp", parIncrCorePD=list(rate=2),
suppLeftPD="runif",parSuppLeftPD=list(min=0,max=0.6),
suppRightPD="runif", parSuppRightPD=list(min=0,max=0.6),
type="trapezoidal")


list2<-SimulateSample(20,originalPD="rnorm",parOriginalPD=list(mean=0,sd=1),
incrCorePD="rexp", parIncrCorePD=list(rate=2),
suppLeftPD="runif",parSuppLeftPD=list(min=0,max=0.6),
suppRightPD="runif", parSuppRightPD=list(min=0,max=0.6),
type="trapezoidal")

# apply the Kolmogorov-Smirnov two-sample test for two different samples
# with the average statistics approach

EpistemicTest(list1$value,list2$value,cutsNumber = 30)
#> [1] 0.8319696

# apply the Kolmogorov-Smirnov two-sample test for two different samples
# with the multi-statistic and antithetic approach

EpistemicTest(list1$value,list2$value,algorithm = "ms",bootstrapMethod = "anti")
#> [1] 0.8319696

# apply the Kolmogorov-Smirnov one-sample test
# with the averaging statistic approach for the first sample

AverageStatisticEpistemicTest(list1$value,sample2=NULL,cutsNumber = 30,y="pnorm")
#> [1] 0.6101717

References