THIS IS A DEVELOPMENT VERSION AND IS NOT RELEASED YET.
kottcalibrate, bounds.hint: equality of model.formulae was checked with all.equal() (which was a good idea) but if clauses didn’t use the right construct (i.e. isTRUE(all.equal())). Fixed.
This version involves little user visible changes. It’s purpose is mainly to ensure a safe transition of EVER to the “R 4.x” series.
Under R 4.0.0 or later, old versions of the EVER package need to be re-installed from sources (i.e. local .tar.gz files). Binary distributions of old versions (i.e. local .zip files) are expected not to work under R 4.0.0 or later.
EVER is now also available from GITHUB at the following URL: https://github.com/DiegoZardetto/EVER
EVER has now a brand new website (built with pkgdown) hosted on GITHUB pages at the following URL: https://diegozardetto.github.io/EVER
kottcalibrate, bounds.hint: now equality of model.formulae is checked with all.equal() instead of identical(), so that environment mismatches are neglected. This is desirable when the model.formulae being compared have been generated within the body of different functions, but are equal when deparsed.
NOTE: ON 2014-11-30 EVER 1.2 WAS ARCHIVED BY CRAN. As it can still be read at the following URL https://cran.r-project.org/package=EVER the package was archived for “policy violations”. The CRAN policy that EVER violated is the following: “Packages should not modify the global environment (user’s workspace)”. At the time EVER was archived such a violation did not even pop up as a ‘WARNING’ messages when running R CMD check -–as-cran (which instead only threw out a ‘NOTE’). The implicated assignment to the .GlobalEnv was performed by EVER’s function ‘kottcalibrate’, and it was: (A) Explicitly documented (you can read the help page of function ‘kottcalibrate’, section ‘Calibration Process Diagnostics’). (B) Totally harmless (and useful, in my opinion).
All .Internal calls have been removed.
.onLoad: bug fixed (thanks to Prof Brian Ripley). system.file is now invoked with lib.loc = libname.
population.check: less tight check on factor columns: now ONLY the values stored into df.population and into template are requested to be identical (and equally ordered). Formerly a mismatch in the levels attribute raised an error (which actually does not affect the results).
bounds.hint: works better when either of the bounds of the star interval is not finite
g.range: small change (now works better if zeros in initial and calibrated weights are not tied)
kottcalibrate: minor bug fixed (added drop = FALSE in mm[ , !zz, drop = FALSE])
kottcalibrate: checks added on bounds (must be numeric and must satisfy bounds[1] <= 1 <= bounds[2])
population.check: default value (FALSE) for ‘partition’ argument was lacking, despite documented.
g.range: minor bug (was not working for objects that have undergone subsequent calibrations, due to erroneous name reconstruction of last but one weights)
LICENCE: removed non ASCII characters
kottby, kottby.user, kott.ratio, kott.quantile, kott.regcoef: NEW argument ‘vartype’ allows to request every possible combination of: SE, CV, CV%, Var
NEW bounds.hint function: helps the user when performing range restricted calibration. Suggests a sound ‘bounds’ value for which kottcalibrate is likely to converge.
g.range BUG fix: g-weights lower value was erroneously 0 whenever there was a zero calibration weight corresponding to a non zero direct weight (thus erroneously missing possible negative g-weights).
Typo in a kottcalibrate message (‘prameter’)
Help page for population.check missed a \code{} around a ‘kottcalibrate’ and had 2 ‘deskott’ instead of ‘data’