WBPP on raw vs Calibrated files

In general I'd say no but I don't let PI decide what the weights are nor I use WBPP. YMMV.

I don't use WBPP either.

During my manual process, I do the weighting AFTER calibration/cosmetic correction. If you tell me that WBPP does it before then I am not sure why it would do it at this stage and I don't think it would matter either.

Side note: If your dataset is THAT large you'll run into out of memory issues during stacking and you'll need to first stack groups of X images, then stack the master of each group. With the massive resolution of the 6200MM, I have to resort to that technique when stacking >600-800 images on a 128gb memory computer. Not sure WBPP does that.

If I well understand, WBPP (and the same does SubframeSelector when used standalone) issues a warning in your second run, because WBPP thinks you have not calibrated them. But this is not true, because you calibrated them in your first run, therefore the result should be correct, despite the warning.

Also good Tony's idea to split your huge lot in more sections, and run WBPP individually on each of them without integration. Then run integration separately.

Miguel's idea of stacking several integrated submasters, each coming out of a subset integration, poses the following theoretical problem: "Is integration of integrated sub-masters equivalent to a single-run integrated master?" I would say that, though the general answer is no, it might work well with a large set of good quality sub-exposures. As far as I understand the math the final signal-to-noise ratio should be equivalent, while the outlayer rejection would be less effective.
Probably this could be the easiest solution to implement, after testing.
I would be very interested in your test results!

Ciao, Mau

Thanks for all the replies.

A couple of things:

First, the warning is not that I have not calibrated the files.  This is not a warning in the diagnostics section of WBPP.  Rather it is a warning in the Process Console while WBPP is running that says that it sees I am using non-raw files and that may create issues.

Second, there are a lot of ways to split up the work; that is not much of an issue.  I could, for example, just do each channel individually, using a single sub as a registration reference.  

My question was more to understand why dark subtraction and flat division would cause issues for PI's measurement and weighting system and I was wondering if anyone had knowledge or experience of this.  I would rather not inadvertently introduce issues into the data.

I asked this on the PI forum, but 0 responses in a week.  AstroBin is so much more active!

Just don't use the script.  Nothing magical about it.   Memory constraints usually happens during the image integration process.   Running the process manually you can change the buffer sizes, and counts, to break up the pixel row memory usage.