search

pacificclimate / climdex.pcic

Routines to compute ETCCDI Climdex indices
GNU General Public License v3.0
23 stars 13 forks source link

It seems like WSDI and CSDI do not use bootstrapping... #8

Open tatarinova opened 9 years ago

tatarinova commented 9 years ago

Comparison: climdex vs icclim(https://github.com/cerfacs-globc/icclim/releases - the version 4.0.0 with implemented bootstrapping)

Base period: 1961-1970

year climdex icclim abs(diff) with_bs

The difference was only inside base period.

So, I recomputed WSDI without using bootstrap procedure, and I got this:

year climdex icclim abs(diff) without_bs

(The same was for CSDI.)

That made me think that climdex does not use bootstrapping for these two indices (for TX10p, TX90p, TN10p and TN90p it is ok - the results converge).

jameshiebert commented 9 years ago

So climdex.pcic computes separate quantiles for inside and outside of the base period here.

But WSDI and CSDI appear to only use the out of base quantiles here.

In fact threshold.exceedance.duration.index() appears to only take a single set of thresholds as an argument. @bronaugh could you do me a favor and comment on whether you recall a reason for this or whether I should mark this ticket as a bug?

bronaugh commented 9 years ago

This mimics the behaviour of fclimdex / rclimdex. Perhaps we could reach out to Xuebin to try and better understand the reason for this decision.

David

On Thu, Aug 27, 2015 at 8:53 AM, James Hiebert notifications@github.com wrote:

So climdex.pcic computes separate quantiles for inside and outside of the base period here https://github.com/pacificclimate/climdex.pcic/blob/master/R/climdex.r#L102 .

But WSDI and CSDI appear to only use the out of base quantiles here https://github.com/pacificclimate/climdex.pcic/blob/master/R/climdex.r#L1027 .

In fact threshold.exceedance.duration.index() https://github.com/pacificclimate/climdex.pcic/blob/master/R/climdex.r#L1632 appears to only take a single set of thresholds as an argument. @bronaugh https://github.com/bronaugh could you do me a favor and comment on whether you recall a reason for this or whether I should mark this ticket as a bug?

— Reply to this email directly or view it on GitHub https://github.com/pacificclimate/climdex.pcic/issues/8#issuecomment-135476241 .

jameshiebert commented 9 years ago

Thanks David. We'll try and pull in Xuebin for his thoughts.