The goal of FuzzyResampling, a library written in R, is to provide additional resampling procedures, apart from the classical bootstrap (i.e. Efron’s approach, see (Efron and Tibshirani 1994)), for fuzzy data. In the classical approach, secondary samples are drawing with replacement from the initial sample. Therefore most of these bootstrap samples contain repeated values. Moreover, if the size of the primary sample is small then all secondary samples consist of only a few distinct values, which is a serious disadvantage.
To overcome this problem, special resampling algorithms for fuzzy data were introduced (see (Grzegorzewski, Hryniewicz, and Romaniuk 2019, 2020a, 2020b; Grzegorzewski and Romaniuk 2022; M. Romaniuk and Hryniewicz 2019; Maciej Romaniuk and Grzegorzewski 2023)). These methods randomly create values that are “similar” to values from the initial sample, but not exactly the same. During the creation process, some of the characteristics of the initial fuzzy values are kept (e.g., the value, the width, etc.). It was shown that these algorithms provide serious advantages in some statistical areas (like standard error estimation or hypothesis testing) if they are compared with the classical approach. For detailed information concerning the theoretical foundations and practical applications of these resampling methods please see the above-mentioned references.
The initial sample (in the form of a vector or a matrix) should consist of triangular or trapezoidal fuzzy numbers.
Some additional procedures related to these resampling methods are also provided, like calculation of the Bertoluzza et al.’s distance (aka the mid/spread distance, see (Bertoluzza, Corral, and Salas 1995)), estimation of the p-value of the one-sample bootstrapped test for the mean (see (Lubiano et al. 2016)), and estimation of the standard error or the mean-squared error for the mean (see (Grzegorzewski and Romaniuk 2021)). Additionally, there are procedures which randomly generate trapezoidal fuzzy numbers using some well-known statistical distributions (see (Grzegorzewski, Hryniewicz, and Romaniuk 2020a)).
The following procedures are available in the library:
Resampling procedures:
Random generation of the initial samples:
Applications of the bootstrapped samples:
Calculation of the characteristics of fuzzy numbers:
Additional procedures:
You can install the latest development version of FuzzyResampling with:
library(devtools)
install_github("mroman-ibs/FuzzyResampling")You can install the latest stable version from CRAN with:
install.packages("FuzzyResampling")# set seed
set.seed(12345)
# load library
library(FuzzyResampling)
# prepare some fuzzy numbers
fuzzyValues <- matrix(c(0.25,0.5,1,1.25,0.75,1,1.5,2.2,-1,0,0,2),ncol = 4,byrow = TRUE)
fuzzyValues
#> [,1] [,2] [,3] [,4]
#> [1,] 0.25 0.5 1.0 1.25
#> [2,] 0.75 1.0 1.5 2.20
#> [3,] -1.00 0.0 0.0 2.00
# seed PRNG
set.seed(12345)
# generate the secondary sample using the classical approach
ClassicalBootstrap(fuzzyValues)
#> [,1] [,2] [,3] [,4]
#> [1,] 0.75 1 1.5 2.2
#> [2,] -1.00 0 0.0 2.0
#> [3,] 0.75 1 1.5 2.2
# generate the secondary sample using the VA method
VAMethod(fuzzyValues)
#> [,1] [,2] [,3] [,4]
#> [1,] 0.9141124 0.9179438 1.7262290 1.747542
#> [2,] -0.5303703 0.8901852 0.9132088 1.423582
#> [3,] -0.3356065 -0.3321967 -0.3321967 2.664393
# generate the secondary sample (6 fuzzy numbers) using the d-method
DMethod(fuzzyValues, b = 6)
#> [,1] [,2] [,3] [,4]
#> [1,] 0.75 1.0 1.5 3.50
#> [2,] 0.00 1.0 1.5 1.75
#> [3,] 0.25 0.5 1.0 1.25
#> [4,] 0.75 1.0 1.0 3.00
#> [5,] -0.25 0.0 0.5 1.20
#> [6,] -0.50 0.5 1.0 1.25
# calculate the mid/spread distance between the first value
# (from the first row) and the second one (from the second row)
BertoluzzaDistance(fuzzyValues[1,],fuzzyValues[2,])
#> [1] 0.6204837
# seed PRNG
set.seed(1234)
# calculate the p-value using the classical (i.e. Efron's) bootstrap
# for the one-sample test for the mean
OneSampleCTest(fuzzyValues, mu_0 = c(0,0.5,1,1.5))
#> [1] 0.82
# calculate the p-value using the VA resampling method
OneSampleCTest(fuzzyValues, mu_0 = c(0,0.5,1,1.5),resamplingMethod = "VAMethod")
#> [1] 0.91
# seed PRNG
set.seed(1234)
# calculate the p-value using the classical (i.e. Efron's) bootstrap
# for the two-sample test for the mean
TwoSampleCTest(fuzzyValues, fuzzyValues+0.1)
#> [1] 0.86
# calculate the p-value using the VA resampling method
TwoSampleCTest(fuzzyValues, fuzzyValues+0.1,resamplingMethod = "VAMethod")
#> [1] 0.95
# seed PRNG
set.seed(1234)
# calculate the SE of the mean using the classical (i.e. Efron's) bootstrap
SEResamplingMean(fuzzyValues)
#> $mean
#> [1] 0.0075000 0.5100000 0.8416667 1.8413333
#>
#> $SE
#> [1] 0.05487521
# calculate the SE of the mean using the VA resampling method
SEResamplingMean(fuzzyValues, resamplingMethod = "VAMethod")
#> $mean
#> [1] -0.2846996 0.5985998 0.8490542 1.7328917
#>
#> $SE
#> [1] 0.05504196
# calculate the MSE of the given mean using the classical (i.e. Efron's) bootstrap
SEResamplingMean(fuzzyValues, trueMean = c(0,0.5,1,2))
#> $mean
#> [1] 0.0 0.5 1.0 2.0
#>
#> $SE
#> [1] 0.02721175
# calculate the MSE of the given mean using the VA resampling method
SEResamplingMean(fuzzyValues, resamplingMethod = "VAMethod", trueMean = c(0,0.5,1,2))
#> $mean
#> [1] 0.0 0.5 1.0 2.0
#>
#> $SE
#> [1] 0.03119963
# seed PRNG
set.seed(1234)
# generate 10 trapezoidal fuzzy numbers using the normal and uniform distributions
GeneratorNU(10, 0,1,1,2)
#> [,1] [,2] [,3] [,4]
#> [1,] -2.6303454 -1.52367820 -0.75097427 -0.6034145
#> [2,] -1.3180763 -0.02526413 0.54261591 1.1619891
#> [3,] 0.3017466 0.92539517 1.38911338 2.8236569
#> [4,] -3.6293320 -2.38569362 -1.83839083 -0.8292990
#> [5,] -0.4492152 0.21032515 0.61022090 0.9162188
#> [6,] -1.3085376 -0.30454266 1.26572653 2.2735935
#> [7,] -2.4546266 -1.10043751 -0.37349192 0.6144299
#> [8,] -2.4312725 -1.46129002 -0.28782204 1.2145784
#> [9,] -1.8836547 -1.39579705 0.42769842 0.7769981
#> [10,] -2.4667277 -0.93580809 -0.08268549 1.6140993
# calculate the ambiguity for the whole matrix
CalculateAmbiguity(fuzzyValues)
#> [1] 0.3333333 0.4083333 0.5000000# help concerning the VA method
?VAMethod