Jump to content

Welcome to our forums!

Sign In or Register to gain full access to our forums. By registering with us, you'll be able to discuss, share and private message with other members of our community.

Welcome!

Thanks for stopping by the Weather Forums! Please take the time to register and join our community. Feel free to post or start new topics on anything related to the weather or the climate.


Photo

Problems with Gerrity Scores greater 1

verification forecast gerrity skill score

  • Please log in to reply

#1
anzu

Posted 10 August 2015 - 06:54 AM

anzu

    Forum Newbie

  • Members
  • Pip
  • 1 posts

Hello

For a statistical experiment I am calculating lots of Gerrity Skill Scores for contingency tables derived from Temperature forecasts/observations.

I tried the package for R, and when I use some relative probabilities as baseline that I have derived from climatological data, I sometimes get Scores greater than 1. I am pretty sure this should not happen.

Here is a script that can be run in batch mode R which should reproduce the error:

 

library('verification')
A=matrix(c(2394,481,0,0,0,0,4718,1399,0,0,0,0,1740,1755,0,0,0,0,3202,594,0,0,0,0,1717),nrow=5)
pb=c(0.0632,0.2692,0.1461,0.2338,0.2877)
multi.cont(A,pb)

 

 

I know that the contingency table does not represent the baseline probabilities, but I am still puzzled that this happens.

I also coded the procedure in matlab myself, with the information from the original paper and it reproduces the error.

Is there a condition about the baseline probabilities that I am not aware of?

Any Help is welcome.


  • iFred likes this

#2
Black Hole

Posted 03 September 2015 - 12:14 PM

Black Hole

    Daily Contributor

  • Admin
  • 1154 posts
  • LocationBountiful, Ut 84010 ~4575ft

Hello

For a statistical experiment I am calculating lots of Gerrity Skill Scores for contingency tables derived from Temperature forecasts/observations.

I tried the package for R, and when I use some relative probabilities as baseline that I have derived from climatological data, I sometimes get Scores greater than 1. I am pretty sure this should not happen.

Here is a script that can be run in batch mode R which should reproduce the error:

 

library('verification')
A=matrix(c(2394,481,0,0,0,0,4718,1399,0,0,0,0,1740,1755,0,0,0,0,3202,594,0,0,0,0,1717),nrow=5)
pb=c(0.0632,0.2692,0.1461,0.2338,0.2877)
multi.cont(A,pb)

 

 

I know that the contingency table does not represent the baseline probabilities, but I am still puzzled that this happens.

I also coded the procedure in matlab myself, with the information from the original paper and it reproduces the error.

Is there a condition about the baseline probabilities that I am not aware of?

Any Help is welcome.

I work in matlab, send me that code if you want and I can look. 


  • iFred likes this

BS Atmospheric Science University of Utah May 2015

PhD Candidate Atmospheric Sciences

 

--Emphasis on: Forecasting, Mountain Weather, Numerical Weather Prediction, Data Assimilation

 

Winter 2016/17 Snow:
Nov 17: 3.2", 23: 1.6", 28: 9.2" (14)

Dec 1: .5", 16: 2.5", 25: 13" (16)

Jan 2: 5", 3: 2.4", 4: 7.7", 12: 1", 19: 1.2", 21: 13", 23: 6", 24: 1", 25: 3.7", 26: 2.5" (43.5) 

Feb 11: .5", 23: 6.5", 27: 4.5" (13.5)

Mar 5: 5.5" (5.5)

Apr 8: 2", 9: 1.8" (3.8)
Total: 95.3"

Lowest Temp: 2F






Also tagged with one or more of these keywords: verification, forecast, gerrity skill score