Weekly Report 1Oct2010
From RHESSI Wiki
Contents |
Issues with Calibrated Eventlists for Simulated Data
- The Pixon profiles were producing different values for the C statistics depending on the time interval used and whether correction factors were used for each detector during the fits to the detector profiles. The first part of this problem was tracked down with the help of Richard Schwartz. The livetime of the simulation data was zero for the last two seconds of the eventlist file. This lead to the C statistic calculated by Pixon, as it iterates to the final image, to become negative in some cases. The final C statistic calculated in HESSI throws out times with a livetime of zero so it did not encounter this problem. If the time interval of the image was shortened to exclude the times with a livetime of zero, the results then matched. Richard Schwartz has changed one of the routines used to create simulated data so that the livetime is not set to zero at the ned of the file. I am testing to make sure this solves the problem.
- The visibility forward fit imaging algorithm would fail when trying to create an image for simulated data near disk center. The problem turned out to be that there were no visibility measurements used by the algorithm for this data set for detector nine. If there are no visibilities to fit for a detector it does not generate an image map.A check will be put in to insure that all dtectors used to create an image have valid visibility measurements.
We spent more time looking at the visibilities that each detector used and the criteria to keep or discard a visibility measurement. I am trying to figure out why it looks like some visibilities which fit the data poorly are kept in some cases and discarded in other cases. I'm doing this by changing the criteria for the visibilities (i.e. the minimum and maximum number of roll bins used) in the GUI and seeing which get kept and which are discarded, by comparing the visibility maps.
- I am still working on why the C statistics are different if the correction factors are used or not. Using correction factors of 1.0 for each detector should yield the same results as using no correction factors, but this doesn't seem to be the case. I'm still need to look at this more closely, and make sure it is not tied to the livetime issue. All the necessary images have been created for this and the results need to be analyzed today (Monday 4Oct).
- The final issue with the eventlists, and one which I am still working on, is why do the C statistics increase as a function of an increase in the total counts in the final image. Originally this seemed like it might be a calibration problem. However after discussions with Brian and Richard, I've been told that calibration should not be an issue with the simulated data. All detectors are automatically calibrated to each other when the data is created. I've gotten the original data map and the script used to create the calibrated eventlist files for the original data. I am creating a series of files with different total counts to see how the C statistic changes as a function of this with all other parameters remaining the same. i hope this will give some insight into what is happening.
The new RHESSI Website
A meeting is set for next Thursday to discuss the current state of the new RHESSI site and future plans for adding content. It was delayed last week due to scheduling conflicts for the attendees. I'm working on the agenda for the meeting, including how we plan to handle the updating and consolidation of documentation we have in multiple locations. This is currently an issue because we have similar documentation in multiple places which are being updated by different people or not at all. I have added more links to our old documents to the current site and started to reorganize some pages. The look of the new site will also be discussed at the meeting.
New RHESSI Inventory Sheet
Another RHESSI Inventory sheet was received. I checked all inventory and returned it.
Goals for Next Week
- Check issue with using any correction factors, including 1.0 for all detectors, and there effect on the C statistic.
- Check the visibility routines to see what visibilities are used and which are discarded when a visibility imaging algorithm is used. I plan to do this by putting stops in the routines which generate the visibilities and checking to see which ones are kept at every step of the visibilities being created from the original data.
- Create simulated data eventlist files with different total counts. Measure C statistic as a function of total counts. I'm certain there will be an increase as the total counts increase, the question is how quickly, and of course, why?
- Have meeting on the future direction of the new Website and the plans for current and future documentation
