• Sign In
  1. Issues List

Issue 91: Modify the "Evvie's Filter" algorithm to exclude extreme outliers first

issues
Status:closed
Assigned To:Guest
Type:Todo
Area:Topograph
Priority:2
Milestone:1.2
Opened:2011-09-29 22:38 by Evvie Vincow
Changed:2017-10-18 08:33 by Brendan MacLean
Resolved:2011-10-30 13:47 by Nick Shulman
Resolution:Fixed
Closed:2017-10-18 08:33 by Brendan MacLean
2011-09-29 22:38 Evvie Vincow
Title»Modify the "Evvie's Filter" algorithm to exclude extreme outliers first
Assigned To»nicksh@u.washington.edu
Type»Todo
Area»Topograph
Priority»1
Milestone»1.2
Please modify the "Evvie's Filter" algorithm as follows:

1. Determine mean, median, and standard deviation for each cohort at each time point, as before.
2. Exclude all points more than 3 standard deviations above or below the MEDIAN.
3. Determine the *new* mean, median, and SD.
4. Proceed with the rest of the algorithm.
2011-10-06 14:51 Nick Shulman
Evvie's filter is implemented on the HalfLivesForm.
Next, it would be good to have Evvie's filter apply to ResultsByCohort page.
2011-10-06 15:35 Evvie Vincow
Priority1»2
As you pointed out, Nick, the modification did not have as large an impact as I had expected. I'd like to propose one additional step, right after throwing out points > 3 SD above the median:

Throw away any values of 100% or 99% if they are more than 2 SD above the median.
2011-10-30 13:47 Nick Shulman
resolve as Fixed
Statusopen»resolved
Assigned Tonicksh@u.washington.edu»Evvie Vincow
The 99% & 100% part of this filter is now implemented, too.
2017-10-18 08:33 Brendan MacLean
close
Statusresolved»closed
Assigned ToEvvie Vincow»Guest