[Met_help] [rt.rap.ucar.edu #56313] History for Plotting Error Percentiles? (UNCLASSIFIED)

Tressa Fowler via RT met_help at ucar.edu
Thu May 10 09:46:42 MDT 2012


----------------------------------------------------------------
  Initial Request
----------------------------------------------------------------

Classification: UNCLASSIFIED
Caveats: NONE

Are you aware of any easy to use method of plotting the values for the
percentiles of the error such as those output by Point-Stat (E10 - E90) such
that you create a box plot from that data specifically? I've tried doing
this with Excel (stacked column chart), but it's fraught with difficulty due
to the fact that the values always contain negative numbers. 

Thanks.
R/
John

Mr John W. Raby, Meteorologist
U.S. Army Research Laboratory
White Sands Missile Range, NM 88002
(575) 678-2004 DSN 258-2004
FAX (575) 678-1230 DSN 258-1230
Email: john.w.raby2.civ at mail.mil


Classification: UNCLASSIFIED
Caveats: NONE




----------------------------------------------------------------
  Complete Ticket History
----------------------------------------------------------------

Subject: Re: [rt.rap.ucar.edu #56313] Plotting Error Percentiles? (UNCLASSIFIED)
From: Paul Oldenburg
Time: Mon May 07 08:26:38 2012

John,

I am not quite clear on what you are asking for.  Are you interested
in the distribution of the E10 and E90 values in
point_stat output or are you interested in viewing the distribution of
all errors using a box that shows the 10th and
90th percentiles?

If you are interested in the former, you can use the plot_cnt.R script
that John referred you to a couple weeks ago (see
http://www.dtcenter.org/met/users/downloads/analysis_scripts.php).
Use the -column name argument, passing in E10/E90.
Please let us know if you have any questions.

Paul


On 05/07/2012 07:36 AM, Raby, John W USA CIV via RT wrote:
>
> Mon May 07 07:36:40 2012: Request 56313 was acted upon.
> Transaction: Ticket created by john.w.raby2.civ at mail.mil
>         Queue: met_help
>       Subject: Plotting Error Percentiles? (UNCLASSIFIED)
>         Owner: Nobody
>    Requestors: john.w.raby2.civ at mail.mil
>        Status: new
>   Ticket<URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=56313>
>
>
> Classification: UNCLASSIFIED
> Caveats: NONE
>
> Are you aware of any easy to use method of plotting the values for
the
> percentiles of the error such as those output by Point-Stat (E10 -
E90) such
> that you create a box plot from that data specifically? I've tried
doing
> this with Excel (stacked column chart), but it's fraught with
difficulty due
> to the fact that the values always contain negative numbers.
>
> Thanks.
> R/
> John
>
> Mr John W. Raby, Meteorologist
> U.S. Army Research Laboratory
> White Sands Missile Range, NM 88002
> (575) 678-2004 DSN 258-2004
> FAX (575) 678-1230 DSN 258-1230
> Email: john.w.raby2.civ at mail.mil
>
>
> Classification: UNCLASSIFIED
> Caveats: NONE
>
>
>


------------------------------------------------
Subject: Plotting Error Percentiles? (UNCLASSIFIED)
From: Raby, John W USA CIV
Time: Mon May 07 08:45:21 2012

Classification: UNCLASSIFIED
Caveats: NONE

Paul -

I'm interested in the latter of those 2 solutions. The box plot must
show the
distribution of all the raw errors which is expressed by the E10 - E90
values.
I've attached an Excel scatter plot rendition of a box plot which
shows the
error distribution, but ideally, showing this as a true box plot would
be
better. The attached graphic shows plots of the values of E10 - E90
for each
of 5 model/domain combinations for one of our case studies.

R/
John

-----Original Message-----
From: Paul Oldenburg via RT [mailto:met_help at ucar.edu]
Sent: Monday, May 07, 2012 8:27 AM
To: Raby, John W CIV (US)
Subject: Re: [rt.rap.ucar.edu #56313] Plotting Error Percentiles?
(UNCLASSIFIED)

John,

I am not quite clear on what you are asking for.  Are you interested
in the
distribution of the E10 and E90 values in point_stat output or are you
interested in viewing the distribution of all errors using a box that
shows
the 10th and 90th percentiles?

If you are interested in the former, you can use the plot_cnt.R script
that
John referred you to a couple weeks ago (see
http://www.dtcenter.org/met/users/downloads/analysis_scripts.php).
Use
the -column name argument, passing in E10/E90.
Please let us know if you have any questions.

Paul


On 05/07/2012 07:36 AM, Raby, John W USA CIV via RT wrote:
>
> Mon May 07 07:36:40 2012: Request 56313 was acted upon.
> Transaction: Ticket created by john.w.raby2.civ at mail.mil
>         Queue: met_help
>       Subject: Plotting Error Percentiles? (UNCLASSIFIED)
>         Owner: Nobody
>    Requestors: john.w.raby2.civ at mail.mil
>        Status: new
>   Ticket<URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=56313>
>
>
> Classification: UNCLASSIFIED
> Caveats: NONE
>
> Are you aware of any easy to use method of plotting the values for
the
> percentiles of the error such as those output by Point-Stat (E10 -
> E90) such that you create a box plot from that data specifically?
I've
> tried doing this with Excel (stacked column chart), but it's fraught
> with difficulty due to the fact that the values always contain
negative
> numbers.
>
> Thanks.
> R/
> John
>
> Mr John W. Raby, Meteorologist
> U.S. Army Research Laboratory
> White Sands Missile Range, NM 88002
> (575) 678-2004 DSN 258-2004
> FAX (575) 678-1230 DSN 258-1230
> Email: john.w.raby2.civ at mail.mil
>
>
> Classification: UNCLASSIFIED
> Caveats: NONE
>
>
>



Classification: UNCLASSIFIED
Caveats: NONE



------------------------------------------------
Subject: Plotting Error Percentiles? (UNCLASSIFIED)
From: Raby, John W USA CIV
Time: Mon May 07 12:41:06 2012

Classification: UNCLASSIFIED
Caveats: NONE


Paul -

I was able to construct some box plots using Excel. See attached. Not
easy to
do, but I think they are accurate in their usage of the percentiles
from
Point-Stat. I plotted the E10 - E90 values on the y-axis using a lines
type
plot. The final step would be to erase the lines leaving the box plot.
This is
what I would like to do
If there were a way to do it faster and easier.

R/
John

-----Original Message-----
From: Paul Oldenburg via RT [mailto:met_help at ucar.edu]
Sent: Monday, May 07, 2012 8:27 AM
To: Raby, John W CIV (US)
Subject: Re: [rt.rap.ucar.edu #56313] Plotting Error Percentiles?
(UNCLASSIFIED)

John,

I am not quite clear on what you are asking for.  Are you interested
in the
distribution of the E10 and E90 values in point_stat output or are you
interested in viewing the distribution of all errors using a box that
shows
the 10th and 90th percentiles?

If you are interested in the former, you can use the plot_cnt.R script
that
John referred you to a couple weeks ago (see
http://www.dtcenter.org/met/users/downloads/analysis_scripts.php).
Use
the -column name argument, passing in E10/E90.
Please let us know if you have any questions.

Paul


On 05/07/2012 07:36 AM, Raby, John W USA CIV via RT wrote:
>
> Mon May 07 07:36:40 2012: Request 56313 was acted upon.
> Transaction: Ticket created by john.w.raby2.civ at mail.mil
>         Queue: met_help
>       Subject: Plotting Error Percentiles? (UNCLASSIFIED)
>         Owner: Nobody
>    Requestors: john.w.raby2.civ at mail.mil
>        Status: new
>   Ticket<URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=56313>
>
>
> Classification: UNCLASSIFIED
> Caveats: NONE
>
> Are you aware of any easy to use method of plotting the values for
the
> percentiles of the error such as those output by Point-Stat (E10 -
> E90) such that you create a box plot from that data specifically?
I've
> tried doing this with Excel (stacked column chart), but it's fraught
> with difficulty due to the fact that the values always contain
negative
> numbers.
>
> Thanks.
> R/
> John
>
> Mr John W. Raby, Meteorologist
> U.S. Army Research Laboratory
> White Sands Missile Range, NM 88002
> (575) 678-2004 DSN 258-2004
> FAX (575) 678-1230 DSN 258-1230
> Email: john.w.raby2.civ at mail.mil
>
>
> Classification: UNCLASSIFIED
> Caveats: NONE
>
>
>




Classification: UNCLASSIFIED
Caveats: NONE



------------------------------------------------
Subject: Plotting Error Percentiles? (UNCLASSIFIED)
From: Tressa Fowler
Time: Mon May 07 13:35:55 2012

Hi John,

I think you might prefer the R bxp() function. You can pass in the
quantiles to plot directly to this function (sometimes these are
calculated for you by the boxplot() function, but for your 10/90 th
percentiles, you hardwire it). I think this would be easier that the
excel and would eliminate the extraneous lines.

Tressa


On Mon May 07 07:36:40 2012, john.w.raby2.civ at mail.mil wrote:
> Classification: UNCLASSIFIED
> Caveats: NONE
>
> Are you aware of any easy to use method of plotting the values for
the
> percentiles of the error such as those output by Point-Stat (E10 -
E90) such
> that you create a box plot from that data specifically? I've tried
doing
> this with Excel (stacked column chart), but it's fraught with
difficulty due
> to the fact that the values always contain negative numbers.
>
> Thanks.
> R/
> John
>
> Mr John W. Raby, Meteorologist
> U.S. Army Research Laboratory
> White Sands Missile Range, NM 88002
> (575) 678-2004 DSN 258-2004
> FAX (575) 678-1230 DSN 258-1230
> Email: john.w.raby2.civ at mail.mil
>
>
> Classification: UNCLASSIFIED
> Caveats: NONE
>
>



------------------------------------------------
Subject: Plotting Error Percentiles? (UNCLASSIFIED)
From: Raby, John W USA CIV
Time: Mon May 07 13:43:26 2012

Classification: UNCLASSIFIED
Caveats: NONE

Tressa -

Yes, it does sound more preferable. I was wondering if maybe there was
a way
in R, but I'm not very familiar with it other than using the scripts
you
posted on the website which Paul mentioned. Thanks for the tip.

R/
John

-----Original Message-----
From: Tressa Fowler via RT [mailto:met_help at ucar.edu]
Sent: Monday, May 07, 2012 1:36 PM
To: Raby, John W CIV (US)
Subject: [rt.rap.ucar.edu #56313] Plotting Error Percentiles?
(UNCLASSIFIED)

Hi John,

I think you might prefer the R bxp() function. You can pass in the
quantiles
to plot directly to this function (sometimes these are calculated for
you by
the boxplot() function, but for your 10/90 th percentiles, you
hardwire it). I
think this would be easier that the excel and would eliminate the
extraneous
lines.

Tressa


On Mon May 07 07:36:40 2012, john.w.raby2.civ at mail.mil wrote:
> Classification: UNCLASSIFIED
> Caveats: NONE
>
> Are you aware of any easy to use method of plotting the values for
the
> percentiles of the error such as those output by Point-Stat (E10 -
> E90) such that you create a box plot from that data specifically?
I've
> tried doing this with Excel (stacked column chart), but it's fraught
> with difficulty due to the fact that the values always contain
negative
> numbers.
>
> Thanks.
> R/
> John
>
> Mr John W. Raby, Meteorologist
> U.S. Army Research Laboratory
> White Sands Missile Range, NM 88002
> (575) 678-2004 DSN 258-2004
> FAX (575) 678-1230 DSN 258-1230
> Email: john.w.raby2.civ at mail.mil
>
>
> Classification: UNCLASSIFIED
> Caveats: NONE
>
>




Classification: UNCLASSIFIED
Caveats: NONE



------------------------------------------------


More information about the Met_help mailing list