Replicate pearson.test results returned by bestNormalize

[sciabolazza]

I find it difficult to understand how the function bestNormalize calculates the pearson.test to transformations and chooses the best normalized vector. Please consider the following code: ` # create data set.seed(123) library(bestNormalize) library(nortest) x <- rgamma(100, 1, 1)

apply bestNormalize

res <- bestNormalize(x)

chosen transformation is box cox

res$chosen_transform Standardized Box Cox Transformation with 100 nonmissing obs.: Estimated statistics:

• lambda = 0.2547102
• mean (before standardization) = -0.4730509

• sd (before standardization) = 1.00397 res$norm_stats["boxcox"] boxcox 1.002667

  fail to replicate pearson.test results

  p <- pearson.test(res$x.t) p$statistic/p$df P 0.66

  create own norm_stat_fn (equal to that used by bestNormalize)

  my_norm_stat_fn <- function(x) { p <- nortest::pearson.test(x) unname(p$stat/p$df) }

  apply bestNormalize

  res <- bestNormalize(x, norm_stat_fn = my_norm_stat_fn)

  chosen transformation is Yeo-Johnson

  res$chosen_transform Standardized Yeo-Johnson Transformation with 100 nonmissing obs.: Estimated statistics:

• lambda = -0.9410871
• mean (before standardization) = 0.3890851

• sd (before standardization) = 0.2235039 res$norm_stats["yeojohnson"] yeojohnson 0.9466667

  fail to replicate my_norm_stat_fn results

  p <- my_norm_stat_fn(res$x.t); p [1] 1.024

  create own norm_stat_fn (different from that used by bestNormalize)

  my_norm_stat_fn <- function(x) { p <- nortest::pearson.test(x) unname(p$stat) }

  apply bestNormalize

  res <- bestNormalize(x, norm_stat_fn = my_norm_stat_fn)

  chosen transformation is sqrt_x

  res$chosen_transform Standardized sqrt(x a) Transformation with 100 nonmissing obs.: Relevant statistics:

• a = 0
• mean (before standardization) = 0.8387587

• sd (before standardization) = 0.4479005 res$norm_stats["sqrt_x"] sqrt_x 3.344

  fail to replicate my_norm_stat_fn results

  p <- my_norm_stat_fn(res$x.t); p [1] 9.46

  create own norm_stat_fn: assign same value to all transformations

  my_norm_stat_fn <- function(x) { p <- 2 return(p) }

  apply bestNormalize

  res <- bestNormalize(x, norm_stat_fn = my_norm_stat_fn)

  chosen transformation is asinh

  res$chosen_transform Standardized asinh(x) Transformation with 100 nonmissing obs.: Relevant statistics:

• mean (before standardization) = 0.696571

• sd (before standardization) = 0.5410448

  replicate my_norm_stat_fn results

  p <- my_norm_stat_fn(res$x.t); p [1] 2

  observe that custom norm_stat_fn works only if the name of its argument is "x"

  my_norm_stat_fn <- function(m) { p <- nortest::pearson.test(m) unname(p$stat) }

  apply bestNormalize

  res <- bestNormalize(x, norm_stat_fn = my_norm_stat_fn) Error in (function (m) : unused argument (x = c(-0.742466592984817, -0.563796587066655, -1.31311589796953, -0.602562318313787, 3.15457570509193, -0.419595958433174, -0.153341073734975, -0.680628813365879, -0.465967563440651, 1.69922890785271)) `

[gaborcsardi]

Hi, this is a read only mirror of CRAN, please see the package authors and potentially URLs in the DESCRIPTION file. Thanks!