Replies: 9 comments
-
|
From: pc***@fu*** (Patricio Cubillos) Hi Michael, Thanks. Note that there are HST data as well. BART has the smoothing hardcoded into the plotting function. It would be good to set that as a user variable. Thanks for noticing the typo. Best,
|
Beta Was this translation helpful? Give feedback.
-
|
From: mh***@kn*** (Michael Himes) I ran it with almost all of the data points used in Stevenson 2014 (there are 2 data points < 1um that I did not include). The spectrum agrees with Stevenson 2014's for < 5 um. For > 5 um, BART is yielding a smaller eclipse depth by ~0.001. Setting the smoothing via a user variable would probably be a good idea as that is more user-friendly. Michael From: BART-devel <ba***@ph***> on behalf of Patricio Cubillos <pc***@fu***> Hi Michael, Thanks. Note that there are HST data as well. The better would be to run with all available data. BART has the smoothing hardcoded into the plotting function. It would be good to set that as a user variable. Thanks for noticing the typo. Best, On Jun 19, 2017, at 5:47 AM, Michael Himes <mh***@knailto:mh@kn***>> wrote: Hi, I've been running BART in eclipse mode for WASP-12b. The results are generally consistent with the literature. The biggest difference is that the spectrum produced by BART is much smoother than what is in the literature. I am currently re-running it with different Spitzer data (Campo 2011) and 4 additional data points to see if that will change anything. I have attached a previous run’s spectrum; compare that with https://arxiv.org/pdf/1308.0337.pdf or https://arxiv.org/pdf/1209.2412.pdf to see the difference in smoothness. I also noticed that there is a typo in the spectrum plot produced by BART, BART-bestFit-Spectrum.png. The y-axis notes that F_p / F_s is in units of 10^3, but it should be 10^-3. Best, -------------- next part -------------- |
Beta Was this translation helpful? Give feedback.
-
|
From: mh***@kn*** (Michael Himes) Hi, Here's the most recent run of WASP-12b, which includes the 2 data points <1um that I had previously omitted. There is very good agreement with the spectrum in Stevenson2014; see page 6: http://iopscience.iop.org/article/10.1088/0004-637X/791/1/36/pdf . The differences appear to be due to the smoothing in BART. There is also general agreement with Lopez-Morales2010; see blue line in fig5: http://iopscience.iop.org/article/10.1088/2041-8205/716/1/L36/meta So, it looks like BART is getting more or less the correct answer. Michael From: BART-devel <ba***@ph***> on behalf of Michael Himes <mh***@kn***> I ran it with almost all of the data points used in Stevenson 2014 (there are 2 data points < 1um that I did not include). The spectrum agrees with Stevenson 2014's for < 5 um. For > 5 um, BART is yielding a smaller eclipse depth by ~0.001. Setting the smoothing via a user variable would probably be a good idea as that is more user-friendly. Michael From: BART-devel <ba***@ph***> on behalf of Patricio Cubillos <pc***@fu***> Hi Michael, Thanks. Note that there are HST data as well. The better would be to run with all available data. BART has the smoothing hardcoded into the plotting function. It would be good to set that as a user variable. Thanks for noticing the typo. Best, On Jun 19, 2017, at 5:47 AM, Michael Himes <mh***@knailto:mh@kn***>> wrote: Hi, I've been running BART in eclipse mode for WASP-12b. The results are generally consistent with the literature. The biggest difference is that the spectrum produced by BART is much smoother than what is in the literature. I am currently re-running it with different Spitzer data (Campo 2011) and 4 additional data points to see if that will change anything. I have attached a previous run’s spectrum; compare that with https://arxiv.org/pdf/1308.0337.pdf or https://arxiv.org/pdf/1209.2412.pdf to see the difference in smoothness. I also noticed that there is a typo in the spectrum plot produced by BART, BART-bestFit-Spectrum.png. The y-axis notes that F_p / F_s is in units of 10^3, but it should be 10^-3. Best, -------------- next part -------------- |
Beta Was this translation helpful? Give feedback.
-
|
From: pc***@fu*** (Patricio Cubillos) Thanks Michael, Quick question, how does the trace plot look? what’s the acceptance rate? Best,
|
Beta Was this translation helpful? Give feedback.
-
|
From: mh***@kn*** (Michael Himes) The trace plot has one weird area but is otherwise good. See attached. The acceptance rate is ~21%. Michael From: BART-devel <ba***@ph***> on behalf of Patricio Cubillos <pc***@fu***> Thanks Michael, Quick question, how does the trace plot look? what’s the acceptance rate? Best, On Jul 5, 2017, at 8:26 PM, Michael Himes <mh***@knailto:mh@kn***>> wrote: Hi, Here's the most recent run of WASP-12b, which includes the 2 data points <1um that I had previously omitted. There is very good agreement with the spectrum in Stevenson2014; see page 6: http://iopscience.iop.org/article/10.1088/0004-637X/791/1/36/pdf . The differences appear to be due to the smoothing in BART. There is also general agreement with Lopez-Morales2010; see blue line in fig5: http://iopscience.iop.org/article/10.1088/2041-8205/716/1/L36/meta So, it looks like BART is getting more or less the correct answer. Michael From: BART-devel <ba***@phailto:ba@ph***>> on behalf of Michael Himes <mh***@knailto:mh@kn***>> I ran it with almost all of the data points used in Stevenson 2014 (there are 2 data points < 1um that I did not include). The spectrum agrees with Stevenson 2014's for < 5 um. For > 5 um, BART is yielding a smaller eclipse depth by ~0.001. Setting the smoothing via a user variable would probably be a good idea as that is more user-friendly. Michael From: BART-devel <ba***@phailto:ba@ph***>> on behalf of Patricio Cubillos <pc***@fuailto:pc@fu***>> Hi Michael, Thanks. Note that there are HST data as well. The better would be to run with all available data. BART has the smoothing hardcoded into the plotting function. It would be good to set that as a user variable. Thanks for noticing the typo. Best, On Jun 19, 2017, at 5:47 AM, Michael Himes <mh***@knailto:mh@kn***>> wrote: Hi, I've been running BART in eclipse mode for WASP-12b. The results are generally consistent with the literature. The biggest difference is that the spectrum produced by BART is much smoother than what is in the literature. I am currently re-running it with different Spitzer data (Campo 2011) and 4 additional data points to see if that will change anything. I have attached a previous run’s spectrum; compare that with https://arxiv.org/pdf/1308.0337.pdf or https://arxiv.org/pdf/1209.2412.pdf to see the difference in smoothness. I also noticed that there is a typo in the spectrum plot produced by BART, BART-bestFit-Spectrum.png. The y-axis notes that F_p / F_s is in units of 10^3, but it should be 10^-3. Best, <BART-bestFit-Spectrum.png>_______________________________________________ -------------- next part -------------- |
Beta Was this translation helpful? Give feedback.
-
|
From: pc***@fu*** (Patricio Cubillos) OK, that looks fine. Sometimes it happens, it’s like if one of the chains got lost and can’t find the way back to good chi-squares. Seems to happen by chance on some occasions. Also, you might double-check that point at 1 um. The model integrated value is off the model, maybe the filter file does not have the right format? best,
|
Beta Was this translation helpful? Give feedback.
-
|
From: mh***@kn*** (Michael Himes) I noticed that po***@1u*** thought it is just how the data point was expressed since it doesn't show the wavelength coverage, but perhaps not. I used the filter files supplied by some source related to each telescope. They started at 0.2 (Herschel) and 0.3 (ARC) microns respectively. This would require the wavenumber range to reach 50,000 which will cause memory issues. Instead, I cut off the fi***@0.***icrons as they were approximately zero (Herschel) or zero (ARC) at that point. I didn't think this would cause an issue, but maybe it is? What are your thoughts Patricio? Michael From: BART-devel <ba***@ph***> on behalf of Patricio Cubillos <pc***@fu***> OK, that looks fine. Sometimes it happens, it’s like if one of the chains got lost and can’t find the way back to good chi-squares. Seems to happen by chance on some occasions. Also, you might double-check that point at 1 um. The model integrated value is off the model, maybe the filter file does not have the right format? best, On Jul 6, 2017, at 9:49 AM, Michael Himes <mh***@knailto:mh@kn***>> wrote: The trace plot has one weird area but is otherwise good. See attached. The acceptance rate is ~21%. Michael From: BART-devel <ba***@phailto:ba@ph***>> on behalf of Patricio Cubillos <pc***@fuailto:pc@fu***>> Thanks Michael, Quick question, how does the trace plot look? what’s the acceptance rate? Best, On Jul 5, 2017, at 8:26 PM, Michael Himes <mh***@knailto:mh@kn***>> wrote: Hi, Here's the most recent run of WASP-12b, which includes the 2 data points <1um that I had previously omitted. There is very good agreement with the spectrum in Stevenson2014; see page 6: http://iopscience.iop.org/article/10.1088/0004-637X/791/1/36/pdf . The differences appear to be due to the smoothing in BART. There is also general agreement with Lopez-Morales2010; see blue line in fig5: http://iopscience.iop.org/article/10.1088/2041-8205/716/1/L36/meta So, it looks like BART is getting more or less the correct answer. Michael From: BART-devel <ba***@phailto:ba@ph***>> on behalf of Michael Himes <mh***@knailto:mh@kn***>> I ran it with almost all of the data points used in Stevenson 2014 (there are 2 data points < 1um that I did not include). The spectrum agrees with Stevenson 2014's for < 5 um. For > 5 um, BART is yielding a smaller eclipse depth by ~0.001. Setting the smoothing via a user variable would probably be a good idea as that is more user-friendly. Michael From: BART-devel <ba***@phailto:ba@ph***>> on behalf of Patricio Cubillos <pc***@fuailto:pc@fu***>> Hi Michael, Thanks. Note that there are HST data as well. The better would be to run with all available data. BART has the smoothing hardcoded into the plotting function. It would be good to set that as a user variable. Thanks for noticing the typo. Best, On Jun 19, 2017, at 5:47 AM, Michael Himes <mh***@knailto:mh@kn***>> wrote: Hi, I've been running BART in eclipse mode for WASP-12b. The results are generally consistent with the literature. The biggest difference is that the spectrum produced by BART is much smoother than what is in the literature. I am currently re-running it with different Spitzer data (Campo 2011) and 4 additional data points to see if that will change anything. I have attached a previous run’s spectrum; compare that with https://arxiv.org/pdf/1308.0337.pdf or https://arxiv.org/pdf/1209.2412.pdf to see the difference in smoothness. I also noticed that there is a typo in the spectrum plot produced by BART, BART-bestFit-Spectrum.png. The y-axis notes that F_p / F_s is in units of 10^3, but it should be 10^-3. Best, <BART-bestFit-Spectrum.png>_______________________________________________ <output_trace.png>_______________________________________________ -------------- next part -------------- |
Beta Was this translation helpful? Give feedback.
-
|
From: pc***@fu*** (Patricio Cubillos) Must be some small detail in the transmission filter file, otherwise the black dot should be near the blue curve. p
|
Beta Was this translation helpful? Give feedback.
-
|
From: mh***@kn*** (Michael Himes) Ah, good catch--when I converted the filter file to the right units I forgot to reverse the wavelength so that it's increasing. The spectrum looks the same as before, it doesn't appear to have changed things much aside from fixing the po***@1u***ttached is a plot of the spectrum without smoothing. It is still much smoother than what I see in the literature. Is that okay, or should I run it with a smaller wavenumber sampling interval? Michael From: BART-devel <ba***@ph***> on behalf of Patricio Cubillos <pc***@fu***> Must be some small detail in the transmission filter file, otherwise the black dot should be near the blue curve. p On Jul 6, 2017, at 12:29 PM, Michael Himes <mh***@knailto:mh@kn***>> wrote: I noticed that po***@1u*** thought it is just how the data point was expressed since it doesn't show the wavelength coverage, but perhaps not. I used the filter files supplied by some source related to each telescope. They started at 0.2 (Herschel) and 0.3 (ARC) microns respectively. This would require the wavenumber range to reach 50,000 which will cause memory issues. Instead, I cut off the fi***@0.***icrons as they were approximately zero (Herschel) or zero (ARC) at that point. I didn't think this would cause an issue, but maybe it is? What are your thoughts Patricio? Michael From: BART-devel <ba***@phailto:ba@ph***>> on behalf of Patricio Cubillos <pc***@fuailto:pc@fu***>> OK, that looks fine. Sometimes it happens, it’s like if one of the chains got lost and can’t find the way back to good chi-squares. Seems to happen by chance on some occasions. Also, you might double-check that point at 1 um. The model integrated value is off the model, maybe the filter file does not have the right format? best, On Jul 6, 2017, at 9:49 AM, Michael Himes <mh***@knailto:mh@kn***>> wrote: The trace plot has one weird area but is otherwise good. See attached. The acceptance rate is ~21%. Michael From: BART-devel <ba***@phailto:ba@ph***>> on behalf of Patricio Cubillos <pc***@fuailto:pc@fu***>> Thanks Michael, Quick question, how does the trace plot look? what’s the acceptance rate? Best, On Jul 5, 2017, at 8:26 PM, Michael Himes <mh***@knailto:mh@kn***>> wrote: Hi, Here's the most recent run of WASP-12b, which includes the 2 data points <1um that I had previously omitted. There is very good agreement with the spectrum in Stevenson2014; see page 6: http://iopscience.iop.org/article/10.1088/0004-637X/791/1/36/pdf . The differences appear to be due to the smoothing in BART. There is also general agreement with Lopez-Morales2010; see blue line in fig5: http://iopscience.iop.org/article/10.1088/2041-8205/716/1/L36/meta So, it looks like BART is getting more or less the correct answer. Michael From: BART-devel <ba***@phailto:ba@ph***>> on behalf of Michael Himes <mh***@knailto:mh@kn***>> I ran it with almost all of the data points used in Stevenson 2014 (there are 2 data points < 1um that I did not include). The spectrum agrees with Stevenson 2014's for < 5 um. For > 5 um, BART is yielding a smaller eclipse depth by ~0.001. Setting the smoothing via a user variable would probably be a good idea as that is more user-friendly. Michael From: BART-devel <ba***@phailto:ba@ph***>> on behalf of Patricio Cubillos <pc***@fuailto:pc@fu***>> Hi Michael, Thanks. Note that there are HST data as well. The better would be to run with all available data. BART has the smoothing hardcoded into the plotting function. It would be good to set that as a user variable. Thanks for noticing the typo. Best, On Jun 19, 2017, at 5:47 AM, Michael Himes <mh***@knailto:mh@kn***>> wrote: Hi, I've been running BART in eclipse mode for WASP-12b. The results are generally consistent with the literature. The biggest difference is that the spectrum produced by BART is much smoother than what is in the literature. I am currently re-running it with different Spitzer data (Campo 2011) and 4 additional data points to see if that will change anything. I have attached a previous run’s spectrum; compare that with https://arxiv.org/pdf/1308.0337.pdf or https://arxiv.org/pdf/1209.2412.pdf to see the difference in smoothness. I also noticed that there is a typo in the spectrum plot produced by BART, BART-bestFit-Spectrum.png. The y-axis notes that F_p / F_s is in units of 10^3, but it should be 10^-3. Best, <BART-bestFit-Spectrum.png>_______________________________________________ <output_trace.png>_______________________________________________ BART-devel mailing list -------------- next part -------------- |
Beta Was this translation helpful? Give feedback.
-
The following is an archived message from the BART-devel mailing list, which has now closed.
From: mh***@kn*** (Michael Himes)
Date: Mon, 19 Jun 2017 03:47:03 +0000
Subject: [BART-devel] WARP-12b BART run
Hi,
I've been running BART in eclipse mode for WASP-12b. The results are generally consistent with the literature. The biggest difference is that the spectrum produced by BART is much smoother than what is in the literature. I am currently re-running it with different Spitzer data (Campo 2011) and 4 additional data points to see if that will change anything. I have attached a previous run's spectrum; compare that with https://arxiv.org/pdf/1308.0337.pdf or https://arxiv.org/pdf/1209.2412.pdf to see the difference in smoothness.
https://arxiv.org/pdf/1209.2412.pdf
I also noticed that there is a typo in the spectrum plot produced by BART, BART-bestFit-Spectrum.png. The y-axis notes that F_p / F_s is in units of 10^3, but it should be 10^-3.
Best,
Michael
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://physics.ucf.edu/pipermail/bart-devel/attachments/20170619/cd146e0a/attachment-0001.html
-------------- next part --------------
A non-text attachment was scrubbed...
Name: BART-bestFit-Spectrum.png
Type: image/png
Size: 27102 bytes
Desc: BART-bestFit-Spectrum.png
URL: http://physics.ucf.edu/pipermail/bart-devel/attachments/20170619/cd146e0a/attachment-0001.png
Beta Was this translation helpful? Give feedback.
All reactions