invertYJ: Invert a Yeo-Johnson transformation

In akcochrane/ACmisc: Aaron Cochrane's miscellaneous functions

Invert a Yeo-Johnson transformation

Description

Once a Yeo-Johnson transformation has been applied, is may be useful to un-transform the data to recover the raw data. See examples.

Usage

invertYJ(xt, lambda, interval = c(-1e+10, 1e+10), keepMedian = T, keepMAD = T)

Arguments

xt	Transformed values of data
lambda	Lambda value that had transformed data into xt
interval	Interval in which to look for data. If search is difficult, restricting the search to plausible values should help.
keepMedian	Logical. Should the original data's median be kept?
keepMAD	Logical. Should the original data's MAD be kept?

Details

Because YeoJohn defaults to retaining the default median and MAD of the original data, it is not trivial to simply apply the Yeo-Johnson lambda "in reverse." Instead, this function uses 1-dimensional OLS optimization to find the vector of data that has the same median and mean as the transformed data, but also has the lambda applied "in reverse." Such optimization has limits, however; both YeoJohn and invertYJ will work best on relatively small values (e.g., absolute values less than 1000) and on datasets with more than 10 numbers or so.

Examples

x <- 1:10
x1 <- VGAM::yeo.johnson(x,2)
x_recovered <- invertYJ(x1,2)

d <- rnorm(10,3) + rexp(10,4)
d1 <- ACmisc::YeoJohn(d)
invertYJ(attr(d1,'transformedDat'),attr(d1,'lambda'),interval = c(-1e3,1E3)) # fails if there is too large of an interval.

akcochrane/ACmisc documentation built on Nov. 24, 2024, 11:22 a.m.