[b1_ana] A first stab at optimization

Fri May 31 17:34:57 EDT 2013

Good evening,

I've separated out the number of counts per x bin (and the associated 
Azz statistical uncertainty) based on which x bin the counts are getting 
moved into. The file is attached.

In the process of pulling these out, I did find a few problems with the 
rebinning section of the code. Serendipity was with us, though, as there 
were three things wrong which all mostly canceled each other out. With 
the fixed code, the highest x bin doesn't change much at all, and the 
two center ones get slightly smaller, and the lowest x bin gets slightly 
larger. Details can be found here: 
https://hallcweb.jlab.org/wiki/index.php/Elong-13-05-31-b

This doesn't affect the rates or counts before rebinning. If you want to 
compare the fixed re-binned statistics to the uncertainties of each 
spectrometer setting, you can see the plots I showed earlier, the main 
one being 
https://hallcweb.jlab.org/wiki/images/e/ee/2013-05-31-spec-errors.png

Take care,
Ellie

On Fri 31 May 2013 02:03:16 PM EDT, O. A. Rondon wrote:
>
> Hi Elena,
>
> This is an important clarification. Since the data for each x point are
> taken at different times, we would have statistics for each x point
> taken at different SHMS kinematics, covering different regions of the bin.
>
> It would be very good to have a table of the statistics for each point
> indicating the contributions of the central and neighboring SHMS
> settings to each x bin.
>
> Basically, each point would be a combination of data taken at the
> central setting, with corresponding systematics, plus data taken at
> adjacent kinematics, with their separate systematics. This seems OK
> since the points seem to have good overlap in Q^2, according to the Q^2
> vs x plot. Showing the x bin edges on that plot would help to visualize
> the recombination.
>
> In any case, aside from the optimization, it's important to modify the
> drift error bands to approximately reflect the multiple cycles per bin.
> For example, for the nominal proposed run plan, most of the data of the
> lowest x bin will be taken over 6 cycles, but some over the 9 cycles of
> the next higher x bin. A safe estimate of the drift errors for that
> point would be the error per cycle/sqrt(6 cycles), and similarly for the
> other points.
>
> Cheers,
>
> Oscar
>
> Long, Elena wrote:
>>
>> Good morning,
>>
>> I made some edits to the optimizer file to tease out each individual 
>> spectrometer setting's uncertainty (instead of the recombined 
>> statistical uncertainties used in Oscar's file). Details and a link 
>> to the updated file can be found here: 
>> https://hallcweb.jlab.org/wiki/index.php/Elong-13-05-31
>>
>> Take care,
>> Ellie
>>
>>
>> Elena Long, Ph.D.
>> Post Doctoral Research Associate
>> University of New Hampshire
>> elena.long at unh.edu<mailto:elena.long at unh.edu>
>> ellie at jlab.org<mailto:ellie at jlab.org>
>> (603) 862-1962
>> http://nuclear.unh.edu/~elong
>>
>> On 05/31/2013 12:30 AM, O. A. Rondon wrote:
>>
>> Here is a first try at an optimizer. It's clear that since the times at
>> each point are different, the number of cycles is not the same for all
>> points. Also, there are only three independent points, since the HMS
>> data is collected in parallel.
>>
>> The initial observation is that the errors for the low x point improves
>> with more shorter cycles, but the high x points get worse. The optimum
>> is probably a combination of shorter cycles at low x, and 20 h cycles at
>> high x. The HMS data would also benefit from this scheme.
>>
>> The table of overhead times in the proposal is somewhat different to the
>> spreadsheet. There may be an issue with the proposal numbers, or I may
>> be missing a detail somewhere in the spreadsheet. In any case, the basic
>> run plan of 1 cycle/day according to the spreadsheet indicates a
>> somewhat shorter overhead time than the proposal.
>>
>> The spreadsheet and a pdf snapshot are here. To use the spreadsheet,
>> change the number of cycles per day in cell B17, and the number of
>> anneals per cycle pair in cell D7. To highlight the cells with
>> calculated values (formulas) and those with constants, press ctrl-F8.
>> http://twist.phys.virginia.edu/~or/b1/optimizer.ods
>> http://twist.phys.virginia.edu/~or/b1/optimize.pdf
>>
>> Feel free to modify the worksheet.
>>
>> Cheers,
>>
>> Oscar
>>
>>
>>
>>
>>
>>
>>
>>
>> _______________________________________________
>> b1_ana mailing list
>> b1_ana at jlab.org<mailto:b1_ana at jlab.org>
>> https://mailman.jlab.org/mailman/listinfo/b1_ana
>>
>>
>>
>>
>>
>> ------------------------------------------------------------------------
>>
>> _______________________________________________
>> b1_ana mailing list
>> b1_ana at jlab.org
>> https://mailman.jlab.org/mailman/listinfo/b1_ana
>
>
>
>
>
>
> _______________________________________________
> b1_ana mailing list
> b1_ana at jlab.org
> https://mailman.jlab.org/mailman/listinfo/b1_ana
> -- 
> Elena Long, Ph.D.
> Post Doctoral Research Associate
> University of New Hampshire
> elena.long at unh.edu
> ellie at jlab.org
> http://nuclear.unh.edu/~elong
> (603) 862-1962
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://mailman.jlab.org/pipermail/b1_ana/attachments/20130531/b0f06ff3/attachment-0001.html 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: counts_by_xbin.ods
Type: application/vnd.oasis.opendocument.spreadsheet
Size: 19587 bytes
Desc: not available
Url : https://mailman.jlab.org/pipermail/b1_ana/attachments/20130531/b0f06ff3/attachment-0001.ods 