Type: Package
Title: Power Analysis for PLS Classification
Version: 0.2.1
Description: It estimates power and sample size for Partial Least Squares-based methods described in Andreella, et al., (2024), <doi:10.48550/arXiv.2403.10289>.
License: GPL-2 | GPL-3 [expanded from: GPL (≥ 2)]
Encoding: UTF-8
LazyData: true
RoxygenNote: 7.3.2
Imports: compositions, FKSUM, nipals, MASS, foreach, parallel, simukde, ks, mvtnorm, pROC, caret
Language: en-US
BugReports: https://github.com/angeella/powerPLS/issues
URL: https://github.com/angeella/powerPLS
Depends: R (≥ 2.10)
NeedsCompilation: no
Packaged: 2025-03-05 18:24:57 UTC; Andreella
Author: Angela Andreella ORCID iD [aut, cre] (Main author)
Maintainer: Angela Andreella <angela.andreella@unitn.it>
Repository: CRAN
Date/Publication: 2025-03-06 00:00:02 UTC

AUC test

Description

Performs permutation-based test based on AUC

Usage

AUCTest(X, Y, nperm = 100, A, randomization = FALSE,
Y.prob = FALSE, eps = 0.01, scaling = 'auto-scaling',
post.transformation = TRUE, cross.validation = FALSE,...)

Arguments

X

data matrix where columns represent the p variables and rows the n observations.

Y

data matrix where columns represent the two classes and rows the n observations.

nperm

number of permutations. Default to 200.

A

number of score components

randomization

Boolean value. Default to FALSE. If TRUE the permutation p-value is computed

Y.prob

Boolean value. Default FALSE. IF TRUE Y is a probability vector

eps

Default 0.01. eps is used when Y.prob = FALSE to transform Y in a probability vector

scaling

Type of scaling, one of c('auto-scaling', 'pareto-scaling', 'mean-centering'). Default 'auto-scaling'.

post.transformation

Boolean value. TRUE if you want to apply post transformation. Default TRUE

cross.validation

Boolean value. Default FALSE. TRUE if you want to compute the observed test statistic by Nested cross-validation

...

additional arguments related to cross.validation. See repeatedCV_test

Value

List with the following objects:

pv

raw p-value. It equals NA if randomization = FALSE

pv_adj

adjusted p-value. It equals NA if randomization = FALSE

test

estimated test statistic

Author(s)

Angela Andreella

References

For the general framework of power analysis for PLS-based methods see:

Andreella, A., Fino, L., Scarpa, B., & Stocchero, M. (2024). Towards a power analysis for PLS-based methods. arXiv preprint https://arxiv.org/abs/2403.10289.

See Also

Other test statistics implemented: mccTest, scoreTest, dQ2Test, sensitivityTest,F1Test, R2Test, specificityTest, FMTest.

Examples

datas <- simulatePilotData(nvar = 30, clus.size = c(5,5),m = 6,nvar_rel = 5,A = 2)
out <- AUCTest(X = datas$X, Y = datas$Y, A = 1)
out

F1 test

Description

Performs permutation-based test based on F1

Usage

F1Test(X, Y, nperm = 200, A, randomization = FALSE,
Y.prob = FALSE, eps = 0.01, scaling = 'auto-scaling',
post.transformation = TRUE,cross.validation = FALSE,...)

Arguments

X

data matrix where columns represent the p variables and rows the n observations.

Y

data matrix where columns represent the two classes and rows the n observations.

nperm

number of permutations. Default to 200.

A

number of score components

randomization

Boolean value. Default to FALSE. If TRUE the permutation p-value is computed

Y.prob

Boolean value. Default FALSE. IF TRUE Y is a probability vector

eps

Default 0.01. eps is used when Y.prob = FALSE to transform Y in a probability vector

scaling

Type of scaling, one of c('auto-scaling', 'pareto-scaling', 'mean-centering'). Default 'auto-scaling'.

post.transformation

Boolean value. TRUE if you want to apply post transformation. Default TRUE

cross.validation

Boolean value. Default FALSE. TRUE if you want to compute the observed test statistic by Nested cross-validation

...

additional arguments related to cross.validation. See repeatedCV_test

Value

List with the following objects:

pv

raw p-value. It equals NA if randomization = FALSE

pv_adj

adjusted p-value. It equals NA if randomization = FALSE

test

estimated test statistic

Author(s)

Angela Andreella

References

For the general framework of power analysis for PLS-based methods see:

Andreella, A., Fino, L., Scarpa, B., & Stocchero, M. (2024). Towards a power analysis for PLS-based methods. arXiv preprint https://arxiv.org/abs/2403.10289.

See Also

Other test statistics implemented: mccTest, scoreTest, dQ2Test, sensitivityTest,AUCTest, R2Test, specificityTest, FMTest.

Examples

datas <- simulatePilotData(nvar = 30, clus.size = c(15,15),m = 6,nvar_rel = 5,A = 1)
out <- F1Test(X = datas$X, Y = datas$Y, A = 1)
out

FM test

Description

Performs permutation-based test based on FM

Usage

FMTest(X, Y, nperm = 200, A, randomization = FALSE,
Y.prob = FALSE, eps = 0.01, scaling = 'auto-scaling',
post.transformation = TRUE,cross.validation = FALSE,...)

Arguments

X

data matrix where columns represent the p variables and rows the n observations.

Y

data matrix where columns represent the two classes and rows the n observations.

nperm

number of permutations. Default to 200.

A

number of score components

randomization

Boolean value. Default to FALSE. If TRUE the permutation p-value is computed

Y.prob

Boolean value. Default FALSE. IF TRUE Y is a probability vector

eps

Default 0.01. eps is used when Y.prob = FALSE to transform Y in a probability vector

scaling

Type of scaling, one of c('auto-scaling', 'pareto-scaling', 'mean-centering'). Default 'auto-scaling'.

post.transformation

Boolean value. TRUE if you want to apply post transformation. Default TRUE

cross.validation

Boolean value. Default FALSE. TRUE if you want to compute the observed test statistic by Nested cross-validation

...

additional arguments related to cross.validation. See repeatedCV_test

Value

List with the following objects:

pv

raw p-value. It equals NA if randomization = FALSE

pv_adj

adjusted p-value. It equals NA if randomization = FALSE

test

estimated test statistic

Author(s)

Angela Andreella

References

For the general framework of power analysis for PLS-based methods see:

Andreella, A., Fino, L., Scarpa, B., & Stocchero, M. (2024). Towards a power analysis for PLS-based methods. arXiv preprint https://arxiv.org/abs/2403.10289.

See Also

Other test statistics implemented: mccTest, scoreTest, dQ2Test, sensitivityTest,AUCTest, R2Test, specificityTest, F1Test.

Examples

datas <- simulatePilotData(nvar = 30, clus.size = c(5,5),m = 6,nvar_rel = 5,A = 1)
out <- FMTest(X = datas$X, Y = datas$Y, A = 1)
out

Iteration Deflation Algorithm

Description

Performs Iteration Deflation Algorithm

Usage

IDA(X, Y, W)

Arguments

X

Data matrix where columns represent the p variables and rows the n observations.

Y

Vector of class probabilities

W

Weight matrix where columns represent the A components and rows the X variables. Computed from computeWT.

Value

Returns a matrix of scores vectors Tscore.

Author(s)

Angela Andreella

References

Stocchero, M., & Paris, D. (2016). Post-transformation of PLS2 (ptPLS2) by orthogonal matrix: a new approach for generating predictive and orthogonal latent variables. Journal of Chemometrics, 30(5), 242-251.

See Also

PLSc, computeWT

PLS classification

Description

Performs Partial Least Squares classification

Usage

PLSc(X, Y, A, scaling = 'auto-scaling', post.transformation = TRUE,
eps = 0.01, Y.prob = FALSE, transformation = 'ilr')

Arguments

X

Data matrix where columns represent the p variables and rows the n observations.

Y

Data matrix where columns represent the two classes and rows the n observations.

A

Number of score components

scaling

Type of scaling, one of c('auto-scaling', 'pareto-scaling', 'mean-centering'). Default to 'auto-scaling'

post.transformation

Boolean value. TRUE if you want to apply post transformation. Default TRUE

eps

Default 0.01. eps is used when Y.prob = FALSE to transform Y in a probability vector

Y.prob

Boolean value. Default FALSE. IF TRUE Y is a probability vector

transformation

Transformation used to map Y in probability data vector. The options are 'ilr' and 'clr'. Default @ilr.

Value

List with the following objects:

W

Matrix of weights

X_loading

Matrix of X loading

Y_loading

Matrix of Y loading

X

Matrix of X data (predictor variables)

Y

Matrix of Y data (dependent variable)

T_score

Matrix of scores

Y_fitted

Fitted Y matrix

B

Matrix regression coefficients

M

Number of orthogonal components if post.transformation=TRUE is applied.

Author(s)

Angela Andreella

References

Stocchero, M., De Nardi, M., & Scarpa, B. (2021). PLS for classification. Chemometrics and Intelligent Laboratory Systems, 216, 104374.

Examples

datas <- simulatePilotData(nvar = 30, clus.size = c(5,5),m = 6,nvar_rel = 5,A = 2)
out <- PLSc(X = datas$X, Y = datas$Y, A = 3)

R2 test

Description

Performs permutation-based test based on R2

Usage

R2Test(X, Y, nperm = 100, A, randomization = FALSE,
Y.prob = FALSE, eps = 0.01, scaling = 'auto-scaling',
post.transformation = TRUE, cross.validation = FALSE, seed = 123, ...)

Arguments

X

data matrix where columns represent the p variables and rows the n observations.

Y

data matrix where columns represent the two classes and rows the n observations.

nperm

number of permutations. Default to 200.

A

number of score components

randomization

Boolean value. Default to FALSE. If TRUE the permutation p-value is computed

Y.prob

Boolean value. Default FALSE. IF TRUE Y is a probability vector

eps

Default 0.01. eps is used when Y.prob = FALSE to transform Y in a probability vector

scaling

Type of scaling, one of c('auto-scaling', 'pareto-scaling', 'mean-centering'). Default 'auto-scaling'.

post.transformation

Boolean value. TRUE if you want to apply post transformation. Default TRUE

cross.validation

Boolean value. Default FALSE. TRUE if you want to compute the observed test statistic by Nested cross-validation

seed

Seed value

...

additional arguments related to cross.validation. See repeatedCV_test

Value

List with the following objects:

pv

raw p-value. It equals NA if randomization = FALSE

pv_adj

adjusted p-value. It equals NA if randomization = FALSE

test

estimated test statistic

Author(s)

Angela Andreella

References

For the general framework of power analysis for PLS-based methods see:

Andreella, A., Fino, L., Scarpa, B., & Stocchero, M. (2024). Towards a power analysis for PLS-based methods. arXiv preprint https://arxiv.org/abs/2403.10289.

See Also

Other test statistics implemented: mccTest, scoreTest, sensitivityTest, specificityTest,AUCTest, dQ2Test, FMTest, F1Test.

Examples

datas <- simulatePilotData(nvar = 30, clus.size = c(5,5),m = 6,nvar_rel = 5,A = 2)
out <- R2Test(X = datas$X, Y = datas$Y, A = 1)
out

Aqueous Humour data

Description

59 post-mortem aqueous humor samples collected from closed and opened sheep eyes

Usage

aqueous_humour

Format

A data frame with 59 rows and 45 variables:

ID

ID observation

group

class membership (C, O)

R1

metabolic values

R2

metabolic values

R3

metabolic values

R4

metabolic values

R5

metabolic values

R6

metabolic values

R7

metabolic values

R8

metabolic values

R9

metabolic values

R10

metabolic values

R11

metabolic values

R12

metabolic values

R13

metabolic values

R14

metabolic values

R15

metabolic values

R16

metabolic values

R17

metabolic values

R18

metabolic values

R19

metabolic values

R20

metabolic values

R21

metabolic values

R22

metabolic values

R23

metabolic values

R24

metabolic values

R25

metabolic values

R26

metabolic values

R27

metabolic values

R28

metabolic values

R29

metabolic values

R30

metabolic values

R31

metabolic values

R32

metabolic values

R33

metabolic values

R34

metabolic values

R35

metabolic values

R36

metabolic values

R37

metabolic values

R38

metabolic values

R39

metabolic values

R40

metabolic values

R41

metabolic values

R42

metabolic values

R43

metabolic values

Author(s)

Angela Andreella angela.andreella@unive.it

References

https://link.springer.com/article/10.1007/s11306-019-1533-2

Power estimation

Description

Estimates power for a given sample size, type I error level and number of score components.

Usage

computePower(X, Y, A, n, seed = 123,
Nsim = 100, nperm = 200, alpha = 0.05,
scaling = 'auto-scaling', test = 'R2',
Y.prob = FALSE, eps = 0.01, post.transformation = TRUE,
fast = FALSE, transformation = 'clr', ncores = NULL)

Arguments

X

Data matrix where columns represent the p variables and rows the n observations.

Y

Data matrix where columns represent the two classes and rows the n observations.

A

Number of score components

n

Sample size

seed

Seed value

Nsim

Number of simulations

nperm

Number of permutations

alpha

Type I error level

scaling

Type of scaling, one of c('auto-scaling', 'pareto-scaling', 'mean-centering'). Default to 'auto-scaling'

test

Type of test statistic, one of c('score', 'mcc', 'R2'). Default to 'R2'.

Y.prob

Boolean value. Default FALSE. IF TRUE Y is a probability vector

eps

Default 0.01. eps is used when Y.prob = FALSE to transform Y in a probability vector.

post.transformation

Boolean value. TRUE if you want to apply post transformation. Default to TRUE

fast

Use the function fk_density from the FKSUM R package for kernel density estimation. Default to FALSE.

transformation

Transformation used to map Y in probability data vector. The options are 'ilr' and 'clr'.

ncores

Number of cores, default NULL.

Value

Returns a matrix of estimated power for each number of components and tests selected.

Author(s)

Angela Andreella

References

For the general framework of power analysis for PLS-based methods see:

Andreella, A., Fino, L., Scarpa, B., & Stocchero, M. (2024). Towards a power analysis for PLS-based methods. arXiv preprint https://arxiv.org/abs/2403.10289.

Examples

## Not run: 
datas <- simulatePilotData(nvar = 10, clus.size = c(5,5),m = 6,nvar_rel = 5,A = 2)
out <- computePower(X = datas$X, Y = datas$Y, A = 3, n = 20, test = 'R2')

## End(Not run)

Sample size estimation

Description

Compute optimal sample size

Usage

computeSampleSize(n, X, Y, A, alpha, beta,
nperm, Nsim, seed, test = 'R2',...)

Arguments

n

Vector of sample sizes to consider

X

Data matrix where columns represent the p variables and rows the n observations.

Y

Data matrix where columns represent the two classes and rows the n observations.

A

Number of score components

alpha

Type I error level. Default to 0.05

beta

Type II error level. Default to 0.2.

nperm

Number of permutations. Default to 100.

Nsim

Number of simulations. Default to 100.

seed

Seed value

test

Type of test, one of c('score', 'mcc', 'R2'). Default to 'R2'.

...

Further parameters.

Value

Returns a data frame that contains the estimated power for each sample size and number of components considered

Author(s)

Angela Andreella

References

For the general framework of power analysis for PLS-based methods see:

Andreella, A., Fino, L., Scarpa, B., & Stocchero, M. (2024). Towards a power analysis for PLS-based methods. arXiv preprint https://arxiv.org/abs/2403.10289.

See Also

computePower

Examples

## Not run: 
datas <- simulatePilotData(nvar = 10, clus.size = c(5,5),m = 6,nvar_rel = 5,A = 2)
out <- computeSampleSize(X = datas$X, Y = datas$Y, A = 2, A = 3, n = 20, test = 'R2')

## End(Not run)

Compute weight and score matrices from PLSc

Description

Compute weight and score matrices for Partial Least Squares classification

Usage

computeWT(X, Y, A)

Arguments

X

Data matrix where columns represent the p variables and rows the n observations.

Y

Data matrix where columns represent the two classes and rows the n observations.

A

Number of score components

Value

List with the following objects:

W

Matrix of weights

T_score

Matrix of Y scores

R

Matrix of Y residuals

Author(s)

Angela Andreella

dQ2 test

Description

Performs permutation-based test based on dQ2

Usage

dQ2Test(X, Y, nperm = 200, A, randomization = FALSE,
Y.prob = FALSE, eps = 0.01, scaling = 'auto-scaling',
post.transformation = TRUE, class = 1, cross.validation = FALSE, ...)

Arguments

X

data matrix where columns represent the p variables and rows the n observations.

Y

data matrix where columns represent the two classes and rows the n observations.

nperm

number of permutations. Default to 200.

A

number of score components

randomization

Boolean value. Default to FALSE. If TRUE the permutation p-value is computed

Y.prob

Boolean value. Default FALSE. IF TRUE Y is a probability vector

eps

Default 0.01. eps is used when Y.prob = FALSE to transform Y in a probability vector

scaling

Type of scaling, one of c('auto-scaling', 'pareto-scaling', 'mean-centering'). Default 'auto-scaling'.

post.transformation

Boolean value. TRUE if you want to apply post transformation. Default TRUE

class

Numeric value. Specifiy the reference class. Default 1

cross.validation

Boolean value. Default FALSE. TRUE if you want to compute the observed test statistic by Nested cross-validation

...

additional arguments related to cross.validation. See repeatedCV_test

Value

List with the following objects:

pv

raw p-value. It equals NA if randomization = FALSE

pv_adj

adjusted p-value. It equals NA if randomization = FALSE

test

estimated test statistic

Author(s)

Angela Andreella

References

For the general framework of power analysis for PLS-based methods see:

Andreella, A., Fino, L., Scarpa, B., & Stocchero, M. (2024). Towards a power analysis for PLS-based methods. arXiv preprint https://arxiv.org/abs/2403.10289.

See Also

Other test statistics implemented: mccTest, scoreTest, sensitivityTest, specificityTest,AUCTest, R2Test, FMTest, F1Test.

Examples

datas <- simulatePilotData(nvar = 30, clus.size = c(5,5),m = 6,nvar_rel = 5,A = 1)
out <- dQ2Test(X = datas$X, Y = datas$Y, A = 1)
out

MCC test

Description

Performs permutation-based test based on Matthews Correlation Coefficient

Usage

mccTest(X, Y, nperm = 200, A, randomization = FALSE,
Y.prob = FALSE, eps = 0.01, scaling = 'auto-scaling',
post.transformation = TRUE, cross.validation = FALSE, seed = 123, ...)

Arguments

X

data matrix where columns represent the p variables and rows the n observations.

Y

data matrix where columns represent the two classes and rows the n observations.

nperm

number of permutations. Default to 200.

A

number of score components

randomization

Boolean value. Default to FALSE. If TRUE the permutation p-value is computed

Y.prob

Boolean value. Default FALSE. IF TRUE Y is a probability vector

eps

Default 0.01. eps is used when Y.prob = FALSE to transform Y in a probability vector

scaling

Type of scaling, one of c('auto-scaling', 'pareto-scaling', 'mean-centering'). Default 'auto-scaling'.

post.transformation

Boolean value. TRUE if you want to apply post transformation. Default TRUE

cross.validation

Boolean value. Default FALSE. TRUE if you want to compute the observed test statistic by nested cross-validation

seed

Seed value

...

additional arguments related to cross.validation. See repeatedCV_test

Value

List with the following objects:

pv

raw p-value. It equals NA if randomization = FALSE

pv_adj

adjusted p-value. It equals NA if randomization = FALSE

test

estimated test statistic

Author(s)

Angela Andreella

References

For the general framework of power analysis for PLS-based methods see:

Andreella, A., Fino, L., Scarpa, B., & Stocchero, M. (2024). Towards a power analysis for PLS-based methods. arXiv preprint https://arxiv.org/abs/2403.10289.

See Also

Other test statistics implemented: AUCTest, scoreTest, dQ2Test, sensitivityTest,AUCTest, R2Test, specificityTest, FMTest.

Examples

datas <- simulatePilotData(nvar = 30, clus.size = c(15,15),m = 6,nvar_rel = 5,A = 1)
out <- mccTest(X = datas$X, Y = datas$Y, A = 1)
out

post transformed PLS

Description

Performs post transformed Partial Least Squares

Usage

ptPLSc(X, Y, W)

Arguments

X

Data matrix where columns represent the p variables and rows the n observations.

Y

Vector of class probabilities

W

Weight matrix where columns represent the A components and rows the X variables.

Value

List with the following objects:

W

Matrix of weights

G

Post transformation matrix

M

Number of orthogonal components

Author(s)

Angela Andreella

References

Stocchero, M., & Paris, D. (2016). Post-transformation of PLS2 (ptPLS2) by orthogonal matrix: a new approach for generating predictive and orthogonal latent variables. Journal of Chemometrics, 30(5), 242-251.

See Also

PLSc, computeWT

Repeated k-Fold Cross-Validation with Custom Test Metrics

Description

This function performs repeated k-fold cross-validation and computes a selected performance metric across all repetitions and folds. It allows for different types of performance tests, such as MCC, sensitivity, specificity, R2, F1, and more.

Usage

repeatedCV_test(
  data,
  labels,
  k_folds = 5,
  repeats = 3,
  A = 1,
  test_type = "mccTest",
  seed = 1234
)

Arguments

data

A data frame or matrix of features (predictor variables).

labels

A vector of class labels corresponding to the rows of data.

k_folds

An integer specifying the number of cross-validation folds (default = 5).

repeats

An integer specifying the number of times the cross-validation is repeated (default = 3).

A

number of score components

test_type

A character string specifying the type of test to use. Options include:

  • 'mccTest' for Matthews Correlation Coefficient (MCC),

  • 'sensitivityTest' for Sensitivity,

  • 'specificityTest' for Specificity,

  • 'R2Test' for R-squared,

  • 'scoreTest' for Score,

  • 'F1Test' for F1 Score,

  • 'FMTest' for Fowlkes-Mallows Index (FM),

  • 'AUCTest' for Area Under the Curve (AUC),

  • 'dQ2Test' for dQ2.

Default is 'mccTest'.

seed

An integer for setting the random seed to ensure reproducibility (default = 1234).

Value

A numeric value representing the average performance metric across the outer folds.

Examples

datas <- simulatePilotData(nvar = 30, clus.size = c(15,15),m = 6,nvar_rel = 5,A = 1)
data <- datas$X
labels <- datas$Y
mean_mcc <- repeatedCV_test(data, labels, A = 1, test_type = 'mccTest')
cat('Mean MCC:', mean_mcc, '\n')

mean_score <- repeatedCV_test(data, labels, A = 1, test_type = 'scoreTest')
cat('Mean Sensitivity:', mean_score, '\n')

Score test

Description

Performs permutation-based test based on predictive score vector

Usage

scoreTest(X, Y, nperm = 200, A, randomization = FALSE,
Y.prob = FALSE, eps = 0.01, scaling = 'auto-scaling',
post.transformation = TRUE, cross.validation = FALSE, seed = 123, ...)

Arguments

X

data matrix where columns represent the p variables and rows the n observations.

Y

data matrix where columns represent the two classes and rows the n observations.

nperm

number of permutations. Default to 200.

A

number of score components

randomization

Boolean value. Default to FALSE. If TRUE the permutation p-value is computed

Y.prob

Boolean value. Default FALSE. IF TRUE Y is a probability vector

eps

Default 0.01. eps is used when Y.prob = FALSE to transform Y in a probability vector

scaling

Type of scaling, one of c('auto-scaling', 'pareto-scaling', 'mean-centering'). Default 'auto-scaling'.

post.transformation

Boolean value. TRUE if you want to apply post transformation. Default TRUE

cross.validation

Boolean value. Default FALSE. TRUE if you want to compute the observed test statistic by Nested cross-validation

seed

Seed value

...

additional arguments related to cross.validation. See repeatedCV_test

Value

List with the following objects:

pv

raw p-value. It equals NA if randomization = FALSE

pv_adj

adjusted p-value. It equals NA if randomization = FALSE

test

estimated test statistic

Author(s)

Angela Andreella

References

For the general framework of power analysis for PLS-based methods see:

Andreella, A., Fino, L., Scarpa, B., & Stocchero, M. (2024). Towards a power analysis for PLS-based methods. arXiv preprint https://arxiv.org/abs/2403.10289.

See Also

Other test statistics implemented: mccTest, R2Test, sensitivityTest, specificityTest,AUCTest, dQ2Test, FMTest, F1Test.

Examples

datas <- simulatePilotData(nvar = 30, clus.size = c(5,5),m = 6,nvar_rel = 5,A = 2)
out <- scoreTest(X = datas$X, Y = datas$Y, A = 1)
out

sensitivity test

Description

Performs permutation-based test based on sensitivity

Usage

sensitivityTest(X, Y, nperm = 200, A, randomization = FALSE,
Y.prob = FALSE, eps = 0.01, scaling = 'auto-scaling',
post.transformation = TRUE, cross.validation = FALSE, ...)

Arguments

X

data matrix where columns represent the p variables and rows the n observations.

Y

data matrix where columns represent the two classes and rows the n observations.

nperm

number of permutations. Default to 200.

A

number of score components

randomization

Boolean value. Default to FALSE. If TRUE the permutation p-value is computed

Y.prob

Boolean value. Default FALSE. IF TRUE Y is a probability vector

eps

Default 0.01. eps is used when Y.prob = FALSE to transform Y in a probability vector

scaling

Type of scaling, one of c('auto-scaling', 'pareto-scaling', 'mean-centering'). Default 'auto-scaling'.

post.transformation

Boolean value. TRUE if you want to apply post transformation. Default TRUE

cross.validation

Boolean value. Default FALSE. TRUE if you want to compute the observed test statistic by Nested cross-validation

...

additional arguments related to cross.validation. See repeatedCV_test

Value

List with the following objects:

pv

raw p-value. It equals NA if randomization = FALSE

pv_adj

adjusted p-value. It equals NA if randomization = FALSE

test

estimated test statistic

Author(s)

Angela Andreella

References

For the general framework of power analysis for PLS-based methods see:

Andreella, A., Fino, L., Scarpa, B., & Stocchero, M. (2024). Towards a power analysis for PLS-based methods. arXiv preprint https://arxiv.org/abs/2403.10289.

See Also

Other test statistics implemented: mccTest, scoreTest, dQ2Test, specificityTest,AUCTest, R2Test, FMTest, F1Test.

Examples

datas <- simulatePilotData(nvar = 30, clus.size = c(5,5),m = 6,nvar_rel = 5,A = 1)
out <- sensitivityTest(X = datas$X, Y = datas$Y, A = 1)
out

Simulate pilot data

Description

Simulate data matrix under the alternative hypothesis with n observations by kernel density estimation

Usage

sim_XY(out, n, seed = 123, post.transformation = TRUE, A, fast = FALSE)

Arguments

out

Output from PLSc

n

Number of observations to simulate

seed

Seed value

post.transformation

Boolean value. Default to TRUE, i.e., post transformation is applied in PLSc

A

Number of score components used in PLSc.

fast

Use the function fk_density from the FKSUM R package for kernel density estimation. Default to FALSE.

Value

Returns a list:

Y_H1

dependent variable, matrix with 2 columns and n rows (observations)

X_H1

predictor variables, matrix with n rows (observations) and number of columns equal to out$X (i.e., original dataset)

Author(s)

Angela Andreella

References

For the general framework of power analysis for PLS-based methods see:

Andreella, A., Fino, L., Scarpa, B., & Stocchero, M. (2024). Towards a power analysis for PLS-based methods. arXiv preprint https://arxiv.org/abs/2403.10289.

See Also

PLSc, ptPLSc

Examples

datas <- simulatePilotData(nvar = 10, clus.size = c(5,5),m = 6,nvar_rel = 5,A = 2)
out <- PLSc(X = datas$X, Y = datas$Y, A = 3)
out_sim <- sim_XY(out = out, n = 10, A = 3)

Simulate pilot data

Description

Simulate cluster pilot data

Usage

simulatePilotData(seed = 123, nvar, clus.size, nvar_rel,m, A = 2, S1 = NULL, S2 = NULL)

Arguments

seed

Seed value

nvar

Number of variables

clus.size

Vector of two elements, specifying the size of classes (only two classes are considered)

nvar_rel

Number of variables relevant to predict the dependent variable

m

Effect size of separation between classes

A

Oracle number of score components

S1

Covariance matrix for the first class. Default NULL, i.e., the identity is considered.

S2

Covariance matrix for the second class. DefaultNULL, i.e., the identity is considered.

Author(s)

Angela Andreella @return List with the following objects:

X

matrix of predictor variables with nvar columns and the sum of clus.size values as number of rows.

Y

vector of dependent variable with the sum of clus.size values as length

References

For the general framework of power analysis for PLS-based methods see:

Andreella, A., Fino, L., Scarpa, B., & Stocchero, M. (2024). Towards a power analysis for PLS-based methods. arXiv preprint https://arxiv.org/abs/2403.10289.

Examples

datas <- simulatePilotData(nvar = 10, clus.size = c(5,5),m = 6,nvar_rel = 5,A = 2)

specificity test

Description

Performs permutation-based test based on specificity

Usage

specificityTest(X, Y, nperm = 200, A, randomization = FALSE,
Y.prob = FALSE, eps = 0.01, scaling = 'auto-scaling',
post.transformation = TRUE,cross.validation = FALSE,...)

Arguments

X

data matrix where columns represent the p variables and rows the n observations.

Y

data matrix where columns represent the two classes and rows the n observations.

nperm

number of permutations. Default to 200.

A

number of score components

randomization

Boolean value. Default to FALSE. If TRUE the permutation p-value is computed

Y.prob

Boolean value. Default FALSE. IF TRUE Y is a probability vector

eps

Default 0.01. eps is used when Y.prob = FALSE to transform Y in a probability vector

scaling

Type of scaling, one of c('auto-scaling', 'pareto-scaling', 'mean-centering'). Default 'auto-scaling'.

post.transformation

Boolean value. TRUE if you want to apply post transformation. Default TRUE

cross.validation

Boolean value. Default FALSE. TRUE if you want to compute the observed test statistic by Nested cross-validation

...

additional arguments related to cross.validation. See repeatedCV_test

Value

List with the following objects:

pv

raw p-value. It equals NA if randomization = FALSE

pv_adj

adjusted p-value. It equals NA if randomization = FALSE

test

estimated test statistic

Author(s)

Angela Andreella

References

For the general framework of power analysis for PLS-based methods see:

Andreella, A., Fino, L., Scarpa, B., & Stocchero, M. (2024). Towards a power analysis for PLS-based methods. arXiv preprint https://arxiv.org/abs/2403.10289.

See Also

Other test statistics implemented: mccTest, scoreTest, dQ2Test, sensitivityTest,AUCTest, R2Test, FMTest, F1Test.

Examples

datas <- simulatePilotData(nvar = 30, clus.size = c(5,5),m = 6,nvar_rel = 5,A = 1)
out <- specificityTest(X = datas$X, Y = datas$Y, A = 1)
out

Wheezing data

Description

32 urine samples from children at risk of early-onset asthma and those with transient wheezing.

Usage

wheezing

Format

A data frame with 32 rows and 176 variables

Author(s)

Angela Andreella angela.andreella@unive.it

References

https://onlinelibrary.wiley.com/doi/10.1111/pai.12879