search

PNNL-TES / GlobalC

A topdown and bottom up approach for global C cycle
6 stars 4 forks source link

Implement overlap calculation #62

Closed bpbond closed 4 years ago

bpbond commented 4 years ago

Implement method in toyproblem_overlap.Rmd code by @abigailsnyder see #61

Closes #52

bpbond commented 4 years ago

Example output this produces:

Screen Shot 2020-01-18 at 2 19 03 PM
bpbond commented 4 years ago

Thanks Abigail. @jinshijian you've been tagged to review–please go through code carefully, making sure you understand logic and what's going on, so we can discuss tomorrow morning. Thanks.

bpbond commented 4 years ago

@bpbond is the example output from actual data? In other words, are we actually seeing that a bootstrap distribution for a population's mean is statistically significantly non-normal even for large N?

Talked with you about this in the hallway today. The non-normal ones are values after bootstrap but also going through the multiple chains of computation, so I think unsurprising they get skewed a bit.

I'm also surprised that the left_normal (green) curve does not look smoother? From the code, I can't see a reason for it not being smoother, beyond maybe the by = 5 in line 93. That pointy-top is killing me though.

Yes, shrinking the the step size helps this:

Screen Shot 2020-01-23 at 9 50 09 PM
jinshijian commented 4 years ago

Done the review