Uploaded image for project: 'Data Management'
  1. Data Management
  2. DM-23836

DCR templates have incorrect variance

    Details

    • Story Points:
      1
    • Sprint:
      AP S20-4 (March)
    • Team:
      Alert Production
    • Urgent?:
      No

      Description

      The DCR templates for image differencing incorrectly set the variance using only the variance plane of the first subfilter that is loaded. The final variance plane is too low by a factor equal to the actual number of subfilters, resulting in an incorrect threshold for detecting sources in image differencing.

        Attachments

          Activity

          Hide
          sullivan Ian Sullivan added a comment -

          This is a simple fix, but the bug has been complicating my comparisons of deep vs dcr coadds. It's not a large effect, since the variance of the science image usually dominates over the variance of the coadd, but it roughly translates to having used a 4.69 sigma threshold for the dcr coadds, instead of 5.

          Show
          sullivan Ian Sullivan added a comment - This is a simple fix, but the bug has been complicating my comparisons of deep vs dcr coadds. It's not a large effect, since the variance of the science image usually dominates over the variance of the coadd, but it roughly translates to having used a 4.69 sigma threshold for the dcr coadds, instead of 5.
          Hide
          sullivan Ian Sullivan added a comment -
          Show
          sullivan Ian Sullivan added a comment - Pull request is here: https://github.com/lsst/ip_diffim/pull/150
          Hide
          mrawls Meredith Rawls added a comment -

          Good catch. In my brief review, I am taking it as a given that it is "algorithmically correct" to have the variance be equal to the variance for one sub-band multiplied by the total number of sub-bands. As best as I can tell, your one-line change implements that correctly. I'm impressed you noticed the difference between 5- and 4.69-sigma source detection threshold!

          Show
          mrawls Meredith Rawls added a comment - Good catch. In my brief review, I am taking it as a given that it is "algorithmically correct" to have the variance be equal to the variance for one sub-band multiplied by the total number of sub-bands. As best as I can tell, your one-line change implements that correctly. I'm impressed you noticed the difference between 5- and 4.69-sigma source detection threshold!

            People

            • Assignee:
              sullivan Ian Sullivan
              Reporter:
              sullivan Ian Sullivan
              Reviewers:
              Meredith Rawls
              Watchers:
              Ian Sullivan, Meredith Rawls
            • Votes:
              0 Vote for this issue
              Watchers:
              2 Start watching this issue

              Dates

              • Created:
                Updated:
                Resolved:

                Summary Panel