| MPFITFUN error -- only reading the first data value [message #90708] |
Tue, 31 March 2015 15:45  |
graham kerr
Messages: 5
Registered: March 2015
|
Junior Member |
|
|
Hello everyone,
I am trying to use mpfitfun to fit data observed at multiple wavelengths to a blackbody function, with temperature as the only variable; so I'm trying to find the best fit temperature.
My function is called planck_fit_sot.pro, and is below. When I use mpfitfun the output has clearly only tried to fit the first data point. For a few test runs where I simulated blackbody intensities at multiple wavelengths (100 in total), the fitting routine returns the temperature that I set the first data point to. Also, yfit has only one value (the first), with all the rest '0'.
Does anyone know what (presumably silly) mistake I've made here, and why mpfitfun is not using all the data to fit the function?
cheers,
Graham
_____________________________
mpfitfun procedure where wave_rgb & data_rgb are input and temp_range and start_temp are included as optional input :-
if n_elements(start_temp) eq 0 then start_temp = double(6000.0)
parinfo = {value:0.0, fixed:0, limited:[0,0], limits:[0.0,0.0]}
parinfo[0].value = start_temp
parinfo[0].fixed = 0
if n_elements(temp_range) eq 0 then begin
parinfo[0].limited(*) = 0
endif else begin
parinfo[0].limited(*) = 1
parinfo[0].limits[0] = temp_range[0]
parinfo[0].limits[1] = temp_range[1]
endelse
fit_fn = mpfitfun('planck_fit_sot', wave_rgb, data_rgb, err, $
parinfo = parinfo, double = double,$
maxiter = 2000, bestnorm = bestnorm,$
yfit = yfit, perror = perror, dof = dof,$
status = status, errmsg=errmsg)
_____________________________
planck_fit_sot.pro :-
FUNCTION planck_fit_sot, wave, temp
;Some constants
cc = 2.99792458d10 ;cm/s
hh = 6.62606957d-27 ;erg s
kb = 1.3806488d-16 ;erg/K
wave_cm = wave/1.e8 ;cm
bb_fn = dblarr(n_elements(wave))
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;;;;;;
;;;;;;;;;;; DEFINE THE FUNCTION ;;;;;;;;;
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;;;;;;
;2*h*c^2.0
const1 = double(2*hh*cc*cc)
;h*c/k
const2 = double(hh*cc/kb)/wave_cm
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;;;;;;
bb_fn = const1 / ( wave_cm^5.0 * ( exp( const2]/temp)-1. ) )
bb_fn = bb_fn*1.d-8 ;ergs/s/cm^2/sr/Ang
bb_fn_watts = bb_fn/1.e7 ;W/cm^2/sr/Ang
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;;;;;;
return, bb_fn_watts
end
|
|
|
|
| Re: MPFITFUN error -- only reading the first data value [message #90711 is a reply to message #90708] |
Wed, 01 April 2015 11:14  |
Jeremy Bailin
Messages: 618
Registered: April 2008
|
Senior Member |
|
|
On Tuesday, March 31, 2015 at 5:45:03 PM UTC-5, graham kerr wrote:
> Hello everyone,
>
> I am trying to use mpfitfun to fit data observed at multiple wavelengths to a blackbody function, with temperature as the only variable; so I'm trying to find the best fit temperature.
>
> My function is called planck_fit_sot.pro, and is below. When I use mpfitfun the output has clearly only tried to fit the first data point. For a few test runs where I simulated blackbody intensities at multiple wavelengths (100 in total), the fitting routine returns the temperature that I set the first data point to. Also, yfit has only one value (the first), with all the rest '0'.
>
> Does anyone know what (presumably silly) mistake I've made here, and why mpfitfun is not using all the data to fit the function?
>
> cheers,
> Graham
> _____________________________
>
> mpfitfun procedure where wave_rgb & data_rgb are input and temp_range and start_temp are included as optional input :-
>
> if n_elements(start_temp) eq 0 then start_temp = double(6000.0)
> parinfo = {value:0.0, fixed:0, limited:[0,0], limits:[0.0,0.0]}
> parinfo[0].value = start_temp
> parinfo[0].fixed = 0
> if n_elements(temp_range) eq 0 then begin
> parinfo[0].limited(*) = 0
> endif else begin
> parinfo[0].limited(*) = 1
> parinfo[0].limits[0] = temp_range[0]
> parinfo[0].limits[1] = temp_range[1]
> endelse
>
> fit_fn = mpfitfun('planck_fit_sot', wave_rgb, data_rgb, err, $
> parinfo = parinfo, double = double,$
> maxiter = 2000, bestnorm = bestnorm,$
> yfit = yfit, perror = perror, dof = dof,$
> status = status, errmsg=errmsg)
> _____________________________
>
> planck_fit_sot.pro :-
>
> FUNCTION planck_fit_sot, wave, temp
>
> ;Some constants
> cc = 2.99792458d10 ;cm/s
> hh = 6.62606957d-27 ;erg s
> kb = 1.3806488d-16 ;erg/K
>
> wave_cm = wave/1.e8 ;cm
>
> bb_fn = dblarr(n_elements(wave))
>
> ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;;;;;;
> ;;;;;;;;;;; DEFINE THE FUNCTION ;;;;;;;;;
> ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;;;;;;
>
> ;2*h*c^2.0
> const1 = double(2*hh*cc*cc)
>
> ;h*c/k
> const2 = double(hh*cc/kb)/wave_cm
>
> ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;;;;;;
>
> bb_fn = const1 / ( wave_cm^5.0 * ( exp( const2]/temp)-1. ) )
>
> bb_fn = bb_fn*1.d-8 ;ergs/s/cm^2/sr/Ang
>
> bb_fn_watts = bb_fn/1.e7 ;W/cm^2/sr/Ang
> ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;;;;;;
>
> return, bb_fn_watts
>
> end
As written it won't compile -- I'm guessing the "]" isn't supposed to be here:
bb_fn = const1 / ( wave_cm^5.0 * ( exp( const2]/temp)-1. ) )
Assuming that's fixed, I would speculate that your start_temp variable coming into the function is an array (possibly a 1-element array) instead of a scalar. Try "help, start_temp" to check.
-Jeremy.
|
|
|
|
| Re: MPFITFUN error -- only reading the first data value [message #90714 is a reply to message #90711] |
Thu, 02 April 2015 02:16 |
graham kerr
Messages: 5
Registered: March 2015
|
Junior Member |
|
|
Hi,
Yes, that was a typo (but wasn't in my actual code).
start_temp is a float (or double) not a 1-element array so I don't think that's where the error is unfortunately.
On Wednesday, April 1, 2015 at 7:14:08 PM UTC+1, Jeremy Bailin wrote:
> On Tuesday, March 31, 2015 at 5:45:03 PM UTC-5, graham kerr wrote:
>> Hello everyone,
>>
>> I am trying to use mpfitfun to fit data observed at multiple wavelengths to a blackbody function, with temperature as the only variable; so I'm trying to find the best fit temperature.
>>
>> My function is called planck_fit_sot.pro, and is below. When I use mpfitfun the output has clearly only tried to fit the first data point. For a few test runs where I simulated blackbody intensities at multiple wavelengths (100 in total), the fitting routine returns the temperature that I set the first data point to. Also, yfit has only one value (the first), with all the rest '0'.
>>
>> Does anyone know what (presumably silly) mistake I've made here, and why mpfitfun is not using all the data to fit the function?
>>
>> cheers,
>> Graham
>> _____________________________
>>
>> mpfitfun procedure where wave_rgb & data_rgb are input and temp_range and start_temp are included as optional input :-
>>
>> if n_elements(start_temp) eq 0 then start_temp = double(6000.0)
>> parinfo = {value:0.0, fixed:0, limited:[0,0], limits:[0.0,0.0]}
>> parinfo[0].value = start_temp
>> parinfo[0].fixed = 0
>> if n_elements(temp_range) eq 0 then begin
>> parinfo[0].limited(*) = 0
>> endif else begin
>> parinfo[0].limited(*) = 1
>> parinfo[0].limits[0] = temp_range[0]
>> parinfo[0].limits[1] = temp_range[1]
>> endelse
>>
>> fit_fn = mpfitfun('planck_fit_sot', wave_rgb, data_rgb, err, $
>> parinfo = parinfo, double = double,$
>> maxiter = 2000, bestnorm = bestnorm,$
>> yfit = yfit, perror = perror, dof = dof,$
>> status = status, errmsg=errmsg)
>> _____________________________
>>
>> planck_fit_sot.pro :-
>>
>> FUNCTION planck_fit_sot, wave, temp
>>
>> ;Some constants
>> cc = 2.99792458d10 ;cm/s
>> hh = 6.62606957d-27 ;erg s
>> kb = 1.3806488d-16 ;erg/K
>>
>> wave_cm = wave/1.e8 ;cm
>>
>> bb_fn = dblarr(n_elements(wave))
>>
>> ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;;;;;;
>> ;;;;;;;;;;; DEFINE THE FUNCTION ;;;;;;;;;
>> ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;;;;;;
>>
>> ;2*h*c^2.0
>> const1 = double(2*hh*cc*cc)
>>
>> ;h*c/k
>> const2 = double(hh*cc/kb)/wave_cm
>>
>> ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;;;;;;
>>
>> bb_fn = const1 / ( wave_cm^5.0 * ( exp( const2]/temp)-1. ) )
>>
>> bb_fn = bb_fn*1.d-8 ;ergs/s/cm^2/sr/Ang
>>
>> bb_fn_watts = bb_fn/1.e7 ;W/cm^2/sr/Ang
>> ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;;;;;;
>>
>> return, bb_fn_watts
>>
>> end
>
>
> As written it won't compile -- I'm guessing the "]" isn't supposed to be here:
>
> bb_fn = const1 / ( wave_cm^5.0 * ( exp( const2]/temp)-1. ) )
>
> Assuming that's fixed, I would speculate that your start_temp variable coming into the function is an array (possibly a 1-element array) instead of a scalar. Try "help, start_temp" to check.
>
> -Jeremy.
|
|
|
|
| Re: MPFITFUN error -- only reading the first data value [message #90715 is a reply to message #90714] |
Thu, 02 April 2015 03:02 |
graham kerr
Messages: 5
Registered: March 2015
|
Junior Member |
|
|
So after much staring at code I think I have found my (somewhat daft) mistake!
I think that in mpfitfun (and mpfit & mpfitexpr), the function that you specify ('myfunction') must have the independent variable set as 'x' and the dependent variable set as 'p'.
So, in my case, in planck_fit_sot, I changed the function call from
planck_fit_sot, wave, temp
to
planck_fit_sot, x, p
... and then within the code I changed all the 'wave' to 'x' and 'temp' to 'p[0]'.
This seems to have solved my problem.
best,
Graham
On Thursday, April 2, 2015 at 10:16:52 AM UTC+1, graham kerr wrote:
> Hi,
>
> Yes, that was a typo (but wasn't in my actual code).
>
> start_temp is a float (or double) not a 1-element array so I don't think that's where the error is unfortunately.
>
>
>
>
>
> On Wednesday, April 1, 2015 at 7:14:08 PM UTC+1, Jeremy Bailin wrote:
>> On Tuesday, March 31, 2015 at 5:45:03 PM UTC-5, graham kerr wrote:
>>> Hello everyone,
>>>
>>> I am trying to use mpfitfun to fit data observed at multiple wavelengths to a blackbody function, with temperature as the only variable; so I'm trying to find the best fit temperature.
>>>
>>> My function is called planck_fit_sot.pro, and is below. When I use mpfitfun the output has clearly only tried to fit the first data point. For a few test runs where I simulated blackbody intensities at multiple wavelengths (100 in total), the fitting routine returns the temperature that I set the first data point to. Also, yfit has only one value (the first), with all the rest '0'.
>>>
>>> Does anyone know what (presumably silly) mistake I've made here, and why mpfitfun is not using all the data to fit the function?
>>>
>>> cheers,
>>> Graham
>>> _____________________________
>>>
>>> mpfitfun procedure where wave_rgb & data_rgb are input and temp_range and start_temp are included as optional input :-
>>>
>>> if n_elements(start_temp) eq 0 then start_temp = double(6000.0)
>>> parinfo = {value:0.0, fixed:0, limited:[0,0], limits:[0.0,0.0]}
>>> parinfo[0].value = start_temp
>>> parinfo[0].fixed = 0
>>> if n_elements(temp_range) eq 0 then begin
>>> parinfo[0].limited(*) = 0
>>> endif else begin
>>> parinfo[0].limited(*) = 1
>>> parinfo[0].limits[0] = temp_range[0]
>>> parinfo[0].limits[1] = temp_range[1]
>>> endelse
>>>
>>> fit_fn = mpfitfun('planck_fit_sot', wave_rgb, data_rgb, err, $
>>> parinfo = parinfo, double = double,$
>>> maxiter = 2000, bestnorm = bestnorm,$
>>> yfit = yfit, perror = perror, dof = dof,$
>>> status = status, errmsg=errmsg)
>>> _____________________________
>>>
>>> planck_fit_sot.pro :-
>>>
>>> FUNCTION planck_fit_sot, wave, temp
>>>
>>> ;Some constants
>>> cc = 2.99792458d10 ;cm/s
>>> hh = 6.62606957d-27 ;erg s
>>> kb = 1.3806488d-16 ;erg/K
>>>
>>> wave_cm = wave/1.e8 ;cm
>>>
>>> bb_fn = dblarr(n_elements(wave))
>>>
>>> ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;;;;;;
>>> ;;;;;;;;;;; DEFINE THE FUNCTION ;;;;;;;;;
>>> ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;;;;;;
>>>
>>> ;2*h*c^2.0
>>> const1 = double(2*hh*cc*cc)
>>>
>>> ;h*c/k
>>> const2 = double(hh*cc/kb)/wave_cm
>>>
>>> ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;;;;;;
>>>
>>> bb_fn = const1 / ( wave_cm^5.0 * ( exp( const2]/temp)-1. ) )
>>>
>>> bb_fn = bb_fn*1.d-8 ;ergs/s/cm^2/sr/Ang
>>>
>>> bb_fn_watts = bb_fn/1.e7 ;W/cm^2/sr/Ang
>>> ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;;;;;;
>>>
>>> return, bb_fn_watts
>>>
>>> end
>>
>>
>> As written it won't compile -- I'm guessing the "]" isn't supposed to be here:
>>
>> bb_fn = const1 / ( wave_cm^5.0 * ( exp( const2]/temp)-1. ) )
>>
>> Assuming that's fixed, I would speculate that your start_temp variable coming into the function is an array (possibly a 1-element array) instead of a scalar. Try "help, start_temp" to check.
>>
>> -Jeremy.
|
|
|
|
| Re: MPFITFUN error -- only reading the first data value [message #90738 is a reply to message #90715] |
Wed, 08 April 2015 15:55 |
Craig Markwardt
Messages: 1869
Registered: November 1996
|
Senior Member |
|
|
On Thursday, April 2, 2015 at 6:02:07 AM UTC-4, graham kerr wrote:
> So after much staring at code I think I have found my (somewhat daft) mistake!
>
> I think that in mpfitfun (and mpfit & mpfitexpr), the function that you specify ('myfunction') must have the independent variable set as 'x' and the dependent variable set as 'p'.
>
> So, in my case, in planck_fit_sot, I changed the function call from
>
> planck_fit_sot, wave, temp
>
> to
>
> planck_fit_sot, x, p
>
> ... and then within the code I changed all the 'wave' to 'x' and 'temp' to 'p[0]'.
>
> This seems to have solved my problem.
Huh? Nope.
For MPFITFUN, it doesn't matter what you name the parameters, it's just that the first parameter needs to be the independent variable, and the second parameter needs to be the array function parameters. This is documented at the top of MPFITFUN.PRO
For MPFITEXPR, yes indeed, the independent variable in the expression needs to be X and the parameter array needs to be P. This is also documented.
Craig
|
|
|
|
| Re: MPFITFUN error -- only reading the first data value [message #90748 is a reply to message #90738] |
Thu, 09 April 2015 01:28 |
graham kerr
Messages: 5
Registered: March 2015
|
Junior Member |
|
|
On Wednesday, April 8, 2015 at 11:55:44 PM UTC+1, Craig Markwardt wrote:
> On Thursday, April 2, 2015 at 6:02:07 AM UTC-4, graham kerr wrote:
>> So after much staring at code I think I have found my (somewhat daft) mistake!
>>
>> I think that in mpfitfun (and mpfit & mpfitexpr), the function that you specify ('myfunction') must have the independent variable set as 'x' and the dependent variable set as 'p'.
>>
>> So, in my case, in planck_fit_sot, I changed the function call from
>>
>> planck_fit_sot, wave, temp
>>
>> to
>>
>> planck_fit_sot, x, p
>>
>> ... and then within the code I changed all the 'wave' to 'x' and 'temp' to 'p[0]'.
>>
>> This seems to have solved my problem.
>
> Huh? Nope.
>
> For MPFITFUN, it doesn't matter what you name the parameters, it's just that the first parameter needs to be the independent variable, and the second parameter needs to be the array function parameters. This is documented at the top of MPFITFUN.PRO
>
> For MPFITEXPR, yes indeed, the independent variable in the expression needs to be X and the parameter array needs to be P. This is also documented.
>
> Craig
Oh, that seemed to fix my problem, somehow. Do you know what the issue might have been then?
bes,
Graham
|
|
|
|
| Re: MPFITFUN error -- only reading the first data value [message #90749 is a reply to message #90748] |
Thu, 09 April 2015 02:32 |
Mats Löfdahl
Messages: 263
Registered: January 2012
|
Senior Member |
|
|
Den torsdag 9 april 2015 kl. 10:28:30 UTC+2 skrev graham kerr:
> On Wednesday, April 8, 2015 at 11:55:44 PM UTC+1, Craig Markwardt wrote:
>> On Thursday, April 2, 2015 at 6:02:07 AM UTC-4, graham kerr wrote:
>>> So after much staring at code I think I have found my (somewhat daft) mistake!
>>>
>>> I think that in mpfitfun (and mpfit & mpfitexpr), the function that you specify ('myfunction') must have the independent variable set as 'x' and the dependent variable set as 'p'.
>>>
>>> So, in my case, in planck_fit_sot, I changed the function call from
>>>
>>> planck_fit_sot, wave, temp
>>>
>>> to
>>>
>>> planck_fit_sot, x, p
>>>
>>> ... and then within the code I changed all the 'wave' to 'x' and 'temp' to 'p[0]'.
>>>
>>> This seems to have solved my problem.
>>
>> Huh? Nope.
>>
>> For MPFITFUN, it doesn't matter what you name the parameters, it's just that the first parameter needs to be the independent variable, and the second parameter needs to be the array function parameters. This is documented at the top of MPFITFUN.PRO
>>
>> For MPFITEXPR, yes indeed, the independent variable in the expression needs to be X and the parameter array needs to be P. This is also documented.
>>
>> Craig
>
> Oh, that seemed to fix my problem, somehow. Do you know what the issue might have been then?
Maybe this: "I changed all the [...] 'temp' to 'p[0]'". I think the problem is you did not have 'temp[0]' in the first place.
If temp is a 1-element array, the line
bb_fn = const1 / ( wave_cm^5.0 * ( exp( const2]/temp)-1. ) )
might not give you the array with the dimension of wave_cm that you'd expect, but rather a 1-element array like temp. (Even with the bracket corrected.)
I mean:
IDL> a=[1,2,3]
IDL> help,a
A INT = Array[3]
IDL> help,a*2
<Expression> INT = Array[3]
IDL> help,a*[2]
<Expression> INT = Array[1]
IDL> help,a^[2]
<Expression> INT = Array[1]
|
|
|
|
| Re: MPFITFUN error -- only reading the first data value [message #90750 is a reply to message #90749] |
Thu, 09 April 2015 02:42 |
graham kerr
Messages: 5
Registered: March 2015
|
Junior Member |
|
|
On Thursday, April 9, 2015 at 10:32:56 AM UTC+1, Mats Löfdahl wrote:
> Den torsdag 9 april 2015 kl. 10:28:30 UTC+2 skrev graham kerr:
>> On Wednesday, April 8, 2015 at 11:55:44 PM UTC+1, Craig Markwardt wrote:
>>> On Thursday, April 2, 2015 at 6:02:07 AM UTC-4, graham kerr wrote:
>>>> So after much staring at code I think I have found my (somewhat daft) mistake!
>>>>
>>>> I think that in mpfitfun (and mpfit & mpfitexpr), the function that you specify ('myfunction') must have the independent variable set as 'x' and the dependent variable set as 'p'.
>>>>
>>>> So, in my case, in planck_fit_sot, I changed the function call from
>>>>
>>>> planck_fit_sot, wave, temp
>>>>
>>>> to
>>>>
>>>> planck_fit_sot, x, p
>>>>
>>>> ... and then within the code I changed all the 'wave' to 'x' and 'temp' to 'p[0]'.
>>>>
>>>> This seems to have solved my problem.
>>>
>>> Huh? Nope.
>>>
>>> For MPFITFUN, it doesn't matter what you name the parameters, it's just that the first parameter needs to be the independent variable, and the second parameter needs to be the array function parameters. This is documented at the top of MPFITFUN.PRO
>>>
>>> For MPFITEXPR, yes indeed, the independent variable in the expression needs to be X and the parameter array needs to be P. This is also documented.
>>>
>>> Craig
>>
>> Oh, that seemed to fix my problem, somehow. Do you know what the issue might have been then?
>
> Maybe this: "I changed all the [...] 'temp' to 'p[0]'". I think the problem is you did not have 'temp[0]' in the first place.
>
> If temp is a 1-element array, the line
>
> bb_fn = const1 / ( wave_cm^5.0 * ( exp( const2]/temp)-1. ) )
>
> might not give you the array with the dimension of wave_cm that you'd expect, but rather a 1-element array like temp. (Even with the bracket corrected.)
>
> I mean:
>
> IDL> a=[1,2,3]
> IDL> help,a
> A INT = Array[3]
> IDL> help,a*2
> <Expression> INT = Array[3]
> IDL> help,a*[2]
> <Expression> INT = Array[1]
> IDL> help,a^[2]
> <Expression> INT = Array[1]
Ah, that might explain it! Thanks, i'll give it a try later and let you know if that was the issue.
cheers,
Graham
|
|
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
