Details of the review

Editors comments:

The authors present a novel convergence algorithm for estimating the 4PL model in the context of dose-response analysis. While the submission is accompanied by an R package, the manuscript is a computational/methodological paper and not a statistical software paper. Therefore, we have decided to reject the submission at this early stage already and recommend that the authors target either a computational statistics journal or possibly a field journal in dose-response analysis.

Some further details are provided below.

o Currently the package feels like a rather standalone package which however complements functionality provided by package drc. Fitting the same model with package drc should in principle be possible, but the fitting algorithm might fail. It would thus seem reasonable to leverage some of the classes and methods already provided in package drc when implementing package dr4pl. Interested users should be able to switch seemlessly.

o Also beyond drc it would be useful for future users if the contribution were better integrated into the existing literature/software.

o Making the package available from a standard repository like CRAN would greatly enhance its accesibility and visibility. (That's why JSS requires this in its submission guidelines.)

o The replication material cannot be executed:

-- The function Residual() is called, but not exported in the (submitted version of the) package.

-- Then calling FindInitialParms() does not succeed because not all arguments without default values are specified.

-- The part if(!dr4pl.simul$convergence) { stop("aa") } is unclear if within the loop the proportion of non-convergence cases should be determined.

o Package:

The formula method of dr4pl should check the formula and the obtained model matrices and not just assume that the second column corresponds to the dose.
print() methods should also be implemented for the introduced classes.

Comments (2)