| Re: Time series. [message #80098] |
Fri, 04 May 2012 18:08 |
russell.grew
Messages: 74
Registered: February 2005
|
Member |
|
|
Hi Dave,
My IDL is a little rusty. The crude example below should give you the
idea behind the reverse indices and how to access the values in your
bins.
Cheers.
z = randomn(7, 100) + findgen(100)/20.
hist = histogram(z, max =5., min = 0., binsize = 1., reverse_indices =
r)
bin1 = r[r[0]:r[1] - 1]
bin2 = r[r[1]:r[2] - 1]
bin3 = r[r[2]:r[3] - 1]
bin4 = r[r[3]:r[4] - 1]
bin5 = r[r[4]:r[5] - 1]
results = fltarr(5)
results(0) = mean(z[bin1])
results(1) = mean(z[bin2])
results(2) = mean(z[bin3])
results(3) = mean(z[bin4])
results(4) = mean(z[bin5])
plot, findgen(100)/20, z
oplot, findgen(5) + 0.5, results, psym = 2
|
|
|
|
| Re: Time series. [message #80106 is a reply to message #80098] |
Thu, 03 May 2012 22:50  |
d.poreh
Messages: 406
Registered: October 2007
|
Senior Member |
|
|
On Wednesday, May 2, 2012 4:54:07 PM UTC+2, Chris Torrence wrote:
> On Wednesday, May 2, 2012 7:39:10 AM UTC-6, dave poreh wrote:
>> On Tuesday, May 1, 2012 8:57:45 PM UTC+2, Chris Torrence wrote:
>>> On Monday, April 30, 2012 3:26:44 AM UTC-6, dave poreh wrote:
>>>> On Monday, April 30, 2012 9:31:59 AM UTC+2, Mats Löfdahl wrote:
>>>> > Den måndagen den 30:e april 2012 kl. 09:15:45 UTC+2 skrev dave poreh:
>>>> > > Folks
>>>> > > hi,
>>>> > > I am doing some Time series analysis and i wish to plot the mean like ( http://imageshack.us/content_round.php?page=done&l=img20 7/6577/screenshotat20120430091.png). I mean i want to plot the *BLUE* points in this graph. i have tryed to do with:
>>>> > > Smooth(y, 33)
>>>> > >
>>>> > > but the result is not what i want. Is there any help on Time series expert please?
>>>> > > Cheers,
>>>> > > Dave
>>>> >
>>>> > I think you want to bin your data and then plot the mean y value within each bin vs the mid x point.
>>>> Thanks. Now i am thinking about: http://idlastro.gsfc.nasa.gov/idl_html_help/TS_SMOOTH.html
>>>>
>>>> Cheers,
>>>> Dave :-)
>>>
>>> Hi Dave,
>>>
>>> At the risk of getting flamed, you can use the undocumented "NSUM" keyword to the PLOT function:
>>>
>>> r = randomn(s,1000) + findgen(1000)/300
>>> void = LINFIT(findgen(1000),r,YFIT=yfit)
>>> p = plot(r, 'or', /SYM_FILLED, SYM_SIZE=0.5, NAME='Red data')
>>> p1 = plot(r, 'ob', /SYM_FILLED, /OVERPLOT, NAME='Smoothed', nsum=20, /undoc)
>>> p2 = plot(yfit, 'g3', /OVERPLOT, NAME='Linear fit')
>>> l = legend(POSITION=[0.8,0.3])
>>>
>>> In this case, setting NSUM=20 just does a simple average of every 20 points and only plots a single point. Here's what it looks like:
>>> http://www.flickr.com/photos/79705059@N06/6986758828/
>>>
>>> Cheers,
>>> Chris
>>> ExelisVIS
>> Dear Chris
>> hi
>> How could i get the *blue* number for the graph?
>> Cheers,
>> Dave
>
> Hi Dave,
> If you mean get the actual averaged numbers back out of the graph, you can't. Hence the reason for Ken's comment. That NSUM is purely used for visual results, although it would be easy to replicate the result using a simple binned average.
> Cheers,
> Chris
Hi Chris
Actually it is working pretty good for my data, but i need also fit line like this:
fit=linfit(x,y,yfit=yfit)
and for random data it is fine, but for GPS data, i have in some points several registered data, and the linfit does not work fine, so i was thinking to get the average data and do the fit on them, but looks like i can't. :-(
Thanks anyway,
Cheers,
Dave
PS. If you give me your email i could send you one of my data to look at them.
|
|
|
|
| Re: Time series. [message #80110 is a reply to message #80106] |
Wed, 02 May 2012 07:54 |
chris_torrence@NOSPAM
Messages: 528
Registered: March 2007
|
Senior Member |
|
|
On Wednesday, May 2, 2012 7:39:10 AM UTC-6, dave poreh wrote:
> On Tuesday, May 1, 2012 8:57:45 PM UTC+2, Chris Torrence wrote:
>> On Monday, April 30, 2012 3:26:44 AM UTC-6, dave poreh wrote:
>>> On Monday, April 30, 2012 9:31:59 AM UTC+2, Mats Löfdahl wrote:
>>>> Den måndagen den 30:e april 2012 kl. 09:15:45 UTC+2 skrev dave poreh:
>>>> > Folks
>>>> > hi,
>>>> > I am doing some Time series analysis and i wish to plot the mean like ( http://imageshack.us/content_round.php?page=done&l=img20 7/6577/screenshotat20120430091.png). I mean i want to plot the *BLUE* points in this graph. i have tryed to do with:
>>>> > Smooth(y, 33)
>>>> >
>>>> > but the result is not what i want. Is there any help on Time series expert please?
>>>> > Cheers,
>>>> > Dave
>>>>
>>>> I think you want to bin your data and then plot the mean y value within each bin vs the mid x point.
>>> Thanks. Now i am thinking about: http://idlastro.gsfc.nasa.gov/idl_html_help/TS_SMOOTH.html
>>>
>>> Cheers,
>>> Dave :-)
>>
>> Hi Dave,
>>
>> At the risk of getting flamed, you can use the undocumented "NSUM" keyword to the PLOT function:
>>
>> r = randomn(s,1000) + findgen(1000)/300
>> void = LINFIT(findgen(1000),r,YFIT=yfit)
>> p = plot(r, 'or', /SYM_FILLED, SYM_SIZE=0.5, NAME='Red data')
>> p1 = plot(r, 'ob', /SYM_FILLED, /OVERPLOT, NAME='Smoothed', nsum=20, /undoc)
>> p2 = plot(yfit, 'g3', /OVERPLOT, NAME='Linear fit')
>> l = legend(POSITION=[0.8,0.3])
>>
>> In this case, setting NSUM=20 just does a simple average of every 20 points and only plots a single point. Here's what it looks like:
>> http://www.flickr.com/photos/79705059@N06/6986758828/
>>
>> Cheers,
>> Chris
>> ExelisVIS
> Dear Chris
> hi
> How could i get the *blue* number for the graph?
> Cheers,
> Dave
Hi Dave,
If you mean get the actual averaged numbers back out of the graph, you can't. Hence the reason for Ken's comment. That NSUM is purely used for visual results, although it would be easy to replicate the result using a simple binned average.
Cheers,
Chris
|
|
|
|
| Re: Time series. [message #80113 is a reply to message #80110] |
Wed, 02 May 2012 06:39 |
d.poreh
Messages: 406
Registered: October 2007
|
Senior Member |
|
|
On Tuesday, May 1, 2012 8:57:45 PM UTC+2, Chris Torrence wrote:
> On Monday, April 30, 2012 3:26:44 AM UTC-6, dave poreh wrote:
>> On Monday, April 30, 2012 9:31:59 AM UTC+2, Mats Löfdahl wrote:
>>> Den måndagen den 30:e april 2012 kl. 09:15:45 UTC+2 skrev dave poreh:
>>>> Folks
>>>> hi,
>>>> I am doing some Time series analysis and i wish to plot the mean like ( http://imageshack.us/content_round.php?page=done&l=img20 7/6577/screenshotat20120430091.png). I mean i want to plot the *BLUE* points in this graph. i have tryed to do with:
>>>> Smooth(y, 33)
>>>>
>>>> but the result is not what i want. Is there any help on Time series expert please?
>>>> Cheers,
>>>> Dave
>>>
>>> I think you want to bin your data and then plot the mean y value within each bin vs the mid x point.
>> Thanks. Now i am thinking about: http://idlastro.gsfc.nasa.gov/idl_html_help/TS_SMOOTH.html
>>
>> Cheers,
>> Dave :-)
>
> Hi Dave,
>
> At the risk of getting flamed, you can use the undocumented "NSUM" keyword to the PLOT function:
>
> r = randomn(s,1000) + findgen(1000)/300
> void = LINFIT(findgen(1000),r,YFIT=yfit)
> p = plot(r, 'or', /SYM_FILLED, SYM_SIZE=0.5, NAME='Red data')
> p1 = plot(r, 'ob', /SYM_FILLED, /OVERPLOT, NAME='Smoothed', nsum=20, /undoc)
> p2 = plot(yfit, 'g3', /OVERPLOT, NAME='Linear fit')
> l = legend(POSITION=[0.8,0.3])
>
> In this case, setting NSUM=20 just does a simple average of every 20 points and only plots a single point. Here's what it looks like:
> http://www.flickr.com/photos/79705059@N06/6986758828/
>
> Cheers,
> Chris
> ExelisVIS
Dear Chris
hi
How could i get the *blue* number for the graph?
Cheers,
Dave
|
|
|
|
| Re: Time series. [message #80115 is a reply to message #80113] |
Tue, 01 May 2012 13:33 |
Kenneth P. Bowman
Messages: 585
Registered: May 2000
|
Senior Member |
|
|
In article < 11142809.2049.1335898665136.JavaMail.geo-discussion-forums@y njj16 >,
Chris Torrence <gorthmog@gmail.com> wrote:
> At the risk of getting flamed, you can use the undocumented "NSUM" keyword to
> the PLOT function:
>
> p1 = plot(r, 'ob', /SYM_FILLED, /OVERPLOT, NAME='Smoothed', nsum=20, /undoc)
I had better not catch my students hiding analysis operations inside
a PLOT statement! ;-)
Ken
|
|
|
|
| Re: Time series. [message #80118 is a reply to message #80115] |
Tue, 01 May 2012 11:57 |
chris_torrence@NOSPAM
Messages: 528
Registered: March 2007
|
Senior Member |
|
|
On Monday, April 30, 2012 3:26:44 AM UTC-6, dave poreh wrote:
> On Monday, April 30, 2012 9:31:59 AM UTC+2, Mats Löfdahl wrote:
>> Den måndagen den 30:e april 2012 kl. 09:15:45 UTC+2 skrev dave poreh:
>>> Folks
>>> hi,
>>> I am doing some Time series analysis and i wish to plot the mean like ( http://imageshack.us/content_round.php?page=done&l=img20 7/6577/screenshotat20120430091.png). I mean i want to plot the *BLUE* points in this graph. i have tryed to do with:
>>> Smooth(y, 33)
>>>
>>> but the result is not what i want. Is there any help on Time series expert please?
>>> Cheers,
>>> Dave
>>
>> I think you want to bin your data and then plot the mean y value within each bin vs the mid x point.
> Thanks. Now i am thinking about: http://idlastro.gsfc.nasa.gov/idl_html_help/TS_SMOOTH.html
>
> Cheers,
> Dave :-)
Hi Dave,
At the risk of getting flamed, you can use the undocumented "NSUM" keyword to the PLOT function:
r = randomn(s,1000) + findgen(1000)/300
void = LINFIT(findgen(1000),r,YFIT=yfit)
p = plot(r, 'or', /SYM_FILLED, SYM_SIZE=0.5, NAME='Red data')
p1 = plot(r, 'ob', /SYM_FILLED, /OVERPLOT, NAME='Smoothed', nsum=20, /undoc)
p2 = plot(yfit, 'g3', /OVERPLOT, NAME='Linear fit')
l = legend(POSITION=[0.8,0.3])
In this case, setting NSUM=20 just does a simple average of every 20 points and only plots a single point. Here's what it looks like:
http://www.flickr.com/photos/79705059@N06/6986758828/
Cheers,
Chris
ExelisVIS
|
|
|
|
|
|
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
