qqtest 1.2.0

Background

Previous releases depended upon the lmRob() function from the robust package. This package became orphaned on CRAN, archived, and removed. Consequently, the dependency needed to be addressed for `qqtest.

This was not done in time for R core policy, so qqtest was removed from CRAN.

Release 1.2.0 remedies this and adds some functionality. There are now no dependencies beyond base packages grDevices, graphics, and stats.

Changes

When lineup = TRUE the value returned by qqtest() is now just a string containing an obfuscated location of the true data. It is no longer a list with a single names component trueLoc.

New functionality

• a new argument matchMethod now offers several choices for matching location and scale of the data to the that of the test distribution.
  

• new arguments legend.xy and legend.cex let the user make some adjustment to the location and character size of the automatically constructed legend.

• two new functions hideLocation() and revealLocation() have been added for use with the lineup = TRUE case.

  These can also be used more broadly to obfuscate integers in some range from 1 to n. The result is a string which when evaluated will reveal the integer. The point is not encryption, but rather to prevent easy visual decoding.

• when lineup = TRUE, now qqtest() simply returns the string with the true location of the data obfuscated. The function revealLocation() will return the numerical location of the true data. Unfortunately this slight increase in convenience also makes it slightly easier for a user to cheat and hence invalidate the lineup test.

  (In this case, qqtest() used to return the same in a list with named component trueLoc, so any code depending on the result being a list will need to be updated.)

• when lineup = FALSE, qqtest() now invisibly returns a list of named components x, y, and order giving the horizontal and vertical locations of the points in sorted order, as well as the order vector from the input data.

qqtest 1.1.0

Released on CRAN in April 2015, removed by R core March 2020 because of dependency on robust package orphaned and archived then.

Functionality based on

• R. Wayne Oldford (2016), “Self calibrating quantile-quantile plots”, The American Statistician, 70, pp. 74-90.

  https://doi.org/10.1080/00031305.2015.1090338

Implemented matching of slope and intercept using a high breakdown regression estimator of lmRob() from the package robust.

Also introduces R functionality for the K distribution, as an alternative to the chisq distribution. This is particularly attractive for teaching statistics; see the vignette.