Derrick Turk

This is a good question and the entire functionality of `best.fit` is poorly thought out. In addition to the issue you correctly point out with the naive model comparison done...

aRpsDCA doesn't currently do a good job of guessing initial declines from data like this, which begins with a series of zero rate values. This may be improved in future...

It almost certainly doesn't work any more and AFAICT the only difference between the current license and WTFPL is that the current license doesn't permit evil. Is there a need...

I'm not sure what might be causing that. The first place I'd check is probably to make 100% sure that the version of aRpsDCA is the same in both cases,...

That is just super weird... `nlminb` isn't a randomized algorithm as far as I know, and I don't see any changes to that code in the R `stats` version control...

I'm assuming the different results are in a best-overall-fit based on raw rate/time data (i.e. the `best.fit` function)? Can you run the following script (call it "check.R", but GitHub doesn't...

The "with buildup" part is suggestive. That part of the code has been something of a hotspot for errors. Would you mind running the example script above again, but edited...

For what it's worth, on my machine, I get: 4.0.2: ``` > library(aRpsDCA) > sessionInfo() R version 4.0.2 (2020-06-22) Platform: x86_64-w64-mingw32/x64 (64-bit) Running under: Windows 10 x64 (build 19044) Matrix...

That would seem to imply that the issue lies elsewhere. I think the next step would have to be to verify that the exact same data is being sent to...