search

ltrr-arizona-edu / burnr

🔥 Forest fire history analysis in R.
https://ltrr-arizona-edu.github.io/burnr/
15 stars 8 forks source link

Intervals() output for Weibull percentiles mislabeled #202

Closed chguiterman closed 2 years ago

chguiterman commented 2 years ago

We accidentally swapped the quantiles in print.intervals() on the file "intervals.R" and the 0.875 quantile is wrong.

Line 286 reads: quants <- quantile(x, q = c(0.847, 0.5, 0.125))

It should read: quants <- quantile(x, q = c(0.125, 0.5, 0.875))

brews commented 2 years ago

Cool. Good find!

chguiterman commented 2 years ago

and while we're at it...are we sure that 0.847 is right? Should it really be 0.875? That would give us the inner seventy-fifth percentile of intervals. I think maybe FHAES uses the 95-th percentile (0.25, 0.975) but can't be sure: https://help.fhaes.org/ pg. 28

brews commented 2 years ago

Yeah, I think you're right, @chguiterman. That's a good catch. I changed the order of q in this commit and "0.847" might be a typo from this commit. The quantile() function works properly on interval objects, but the values in print() for intervals are definitely mislabeled. This needs to get fixed.

Would we have shifted the order of q to mimic the order or output of another program? I have a vague memory of us considering what people might already be used to. Our original choice of the inner 75%, again, feels like something we deliberately choose to match earlier tools...?

chguiterman commented 2 years ago

OK @brews I follow now. Yes, I think this was a mix-up that is easy to correct. I noticed the quantiles() function working properly, but we should have these values printed correctly too. Yes, these values match both FHAES and FHX2 -- that's what we were going for back in the days of 2017. See this PR to fix the issue Thanks!

brews commented 2 years ago

Updated the description above so it describes all the problems.