Re: Request: Calculating Accuracy with Forecasted Highs/Lows
From: Sgt. Louis Illuminato (illuminato_at_hotpop.com)
Date: 01/18/05
- Next message: Coby Beck: "Re: NASA GISS:Global Temperature Trends: 2004 Summation"
- Previous message: Sgt. Louis Illuminato: "Re: Landsea quits IPCC over agenda pushing"
- In reply to: tenshou-goji: "Request: Calculating Accuracy with Forecasted Highs/Lows"
- Messages sorted by: [ date ] [ thread ]
Date: Tue, 18 Jan 2005 04:30:12 GMT
"tenshou-goji" <someone@nowhere.org> wrote:
>I could use some advice in figuring ways to calculate the accuracy of a
>weather report dealing with the forecasted High and Low temperatures.
>
>So far, my first idea was to calculate the daily accuracy by subtracting the
>forecasted high/low (FH/FL) from the recorded high/low (RH/RL) to get the
>Difference (DH/DL): RL - FL = DL
>
>Then take the absolute value of the Difference (DH/DL)
OK so far...
>and divide by the
>recorded high/low (RH/RL), sutracting one by it and getting a percentage
>accuracy for the high/low (%H/%L): 1 - [DL]/RL = %L.
BZZT! Danger Will Robinson!
Percentages of temperatures are nonsensical. Leave the answer as
degrees.
Different but also useful information can be obtained if you don't
take the absolute value. Averaged over time this gives you the
forecast "bias"; that is, the tendency of the forecast to be
systematically too high or too low.
Of course you should also do that Google thing as James suggested.
-- [signature missing]
- Next message: Coby Beck: "Re: NASA GISS:Global Temperature Trends: 2004 Summation"
- Previous message: Sgt. Louis Illuminato: "Re: Landsea quits IPCC over agenda pushing"
- In reply to: tenshou-goji: "Request: Calculating Accuracy with Forecasted Highs/Lows"
- Messages sorted by: [ date ] [ thread ]
Relevant Pages
|
|
