| Title: | A Novel Automatic Shifted Log Transformation |
|---|---|
| Description: | A novel parametrization of log transformation and a shift parameter to automate the transformation process are proposed in R package 'AutoTransQF' based on Feng et al. (2016). Please read Feng et al. (2016) <doi:10.1002/sta4.104> for more details of the method. |
| Authors: | Yue Hu [aut, cre], Hyeon Lee [aut], J. S. Marron [aut] |
| Maintainer: | Yue Hu <[email protected]> |
| License: | GPL-3 |
| Version: | 0.1.3 |
| Built: | 2024-11-24 06:32:23 UTC |
| Source: | CRAN |
The R package AutoTransQF based on Feng et al.(2016) introduces a novel parametrization of log transformation and a shift parameter to automate the transformation process. Please read Feng et al. (2016) <doi: 10.1002/sta4.104> for more details of the method.
Yue Hu [aut, cre], Hyeon Lee [aut], J. S. Marron [aut]
The function is used to calculate the Anderson-Darling test statistic of standard normal distribution. The input needs to be vectors with seven or more entries.
## Calculate Anderson-Darling test statistic for vector x: ADStatQF(x)## Calculate Anderson-Darling test statistic for vector x: ADStatQF(x)
x |
a vector with seven or more entries. |
Returns the Anderson-Darling test statistic for the given vector.
Yue Hu, Hyeon Lee, J. S. Marron
Feng, Q. , Hannig J. , Marron, J. S. (2016). A Note on Automatic Data Transformation. STAT, 5, 82-87. doi: 10.1002/sta4.104
## Generate a vector z from normal distribution with mean 0 and standard deviation 6. z = rnorm(10, mean = 0, sd = 6) ## Calculate the Anderson-Darling test statistic for z ADStatQF(z)## Generate a vector z from normal distribution with mean 0 and standard deviation 6. z = rnorm(10, mean = 0, sd = 6) ## Calculate the Anderson-Darling test statistic for z ADStatQF(z)
This function transforms individual vectors into normality. Paper from Feng et al. (2016) includes more details about the transformation mechanism.
autotransfuncQF(vari, istat, paraindex)autotransfuncQF(vari, istat, paraindex)
vari |
a vector needs to be transformed |
istat |
a value representing the type of test statistic for evaluation of normality of the transformed vector. If |
paraindex |
a value delivered to calculate the specific shift parameter beta |
Returns a transformed vector with the shift parameter calculated by the input paraindex
Yue Hu, Hyeon Lee, J. S. Marron
Feng, Q. , Hannig J. , Marron, J. S. (2016). A Note on Automatic Data Transformation. STAT, 5, 82-87. doi: 10.1002/sta4.104
## Generate a vector vec from Gamma distribution with parameters shape 1 and scale 2. vec = rgamma(50, shape = 1, scale = 2) ## Choose Anderson-Darling test statistic for transformed vector. ## Assign paraindex to be 0.9. autotransfuncQF(vec, istat = 1, paraindex = 0.9) ## Choose skewness for transformed vector ## Assign paraindex to be 0.9 autotransfuncQF(vec, istat = 2, paraindex = 0.9)## Generate a vector vec from Gamma distribution with parameters shape 1 and scale 2. vec = rgamma(50, shape = 1, scale = 2) ## Choose Anderson-Darling test statistic for transformed vector. ## Assign paraindex to be 0.9. autotransfuncQF(vec, istat = 1, paraindex = 0.9) ## Choose skewness for transformed vector ## Assign paraindex to be 0.9 autotransfuncQF(vec, istat = 2, paraindex = 0.9)
This function helps to transform each vector of the matrix into normality based on the optimal test statistic of transformed vectors.
## The function tries to transform each vector of mdata into normality AutoTransQF(mdata, paramstruct = list(istat, iscreenwrite, FeatureNames))## The function tries to transform each vector of mdata into normality AutoTransQF(mdata, paramstruct = list(istat, iscreenwrite, FeatureNames))
mdata |
the matrix needs to be transformed. |
paramstruct |
A list with three entries istat, iscreenwrite and FeatureNames respectively. Missing entries will be set to default. |
istat |
a value representing the type of test statistic for evaluation of normality of the transformed vector with default to be |
iscreenwrite |
Whether there is screenwrite with default to be |
FeatureNames |
Contains feature names of each vector with default to be 'Feature1' |
Returns a list with three elements:
data |
the transformed matrix |
beta |
a list of all shift parameters beta |
alpha |
a list of all shift parameters alpha |
When a vector of the original matrix is not transformed, its corresponding alpha and beta are both -1.
Yue Hu, Hyeon Lee, J. S. Marron
Feng, Q. , Hannig J. , Marron, J. S. (2016). A Note on Automatic Data Transformation. STAT, 5, 82-87. doi: 10.1002/sta4.104
## Create a random matrix x. x = matrix(rgamma(40, shape = 1, scale = 2), nrow = 4) ## Transform matrix x in default setting and ## output transformed data AutoTransQF(x)$data ## Transform matrix x in default setting and ## output a list of shift parameter beta AutoTransQF(x)$beta ## Transform matrix x with feature names and ## output a list of shift parameter alpha Names = c('Feature1', 'Feature2', 'Feature3', 'Feature4') AutoTransQF(x, paramstruct = list(FeatureNames = Names))$alpha ## Transform matrix x with feature names, progress to screen, ## and apply standard skewness statistic to transformed vectors AutoTransQF(x, paramstruct = list(istat = 2, iscreenwrite = 1, FeatureNames = Names)) ## Transform matrix x with progress to screen and ## apply standard skewness statistic to transformed vectors AutoTransQF(x, paramstruct = list(istat = 2, iscreenwrite = 1))## Create a random matrix x. x = matrix(rgamma(40, shape = 1, scale = 2), nrow = 4) ## Transform matrix x in default setting and ## output transformed data AutoTransQF(x)$data ## Transform matrix x in default setting and ## output a list of shift parameter beta AutoTransQF(x)$beta ## Transform matrix x with feature names and ## output a list of shift parameter alpha Names = c('Feature1', 'Feature2', 'Feature3', 'Feature4') AutoTransQF(x, paramstruct = list(FeatureNames = Names))$alpha ## Transform matrix x with feature names, progress to screen, ## and apply standard skewness statistic to transformed vectors AutoTransQF(x, paramstruct = list(istat = 2, iscreenwrite = 1, FeatureNames = Names)) ## Transform matrix x with progress to screen and ## apply standard skewness statistic to transformed vectors AutoTransQF(x, paramstruct = list(istat = 2, iscreenwrite = 1))
The dataset from Miedema et al. (2012) is built based on 49 hematoxylin and eosin stained slides of distinctive melanocytic lesions.
MelanomaMelanoma
A data frame with 10152 observations on 49 variables where columns serve as data objects and rows serve as features. All 49 features are numeric variables including Area, Hu.1, Hu.2 etc.
Miedema, Jayson, et al. (2012). Image and statistical analysis of melanocytic histology. Histopathology, 61(3), pp.436-444. doi: 10.1111/j.1365-2559.2012.04229.x