Ticket #60 (closed defect: moved)

Opened 13 years ago

Last modified 6 years ago

Quality of exponential curve fit is questionable

Reported by: curtis Owned by: nor
Priority: critical Milestone:
Component: slim-plotter Severity: serious
Keywords: Cc:
Blocked By: Blocking:

Description (last modified by curtis) (diff)

The single exponential curve fits look good, but the raw chi-squared error values seem strange. Slim Plotter currently employs some tactics to normalize the value, but I do not know enough about  chi-square tests to be sure these operations are the proper way to correct the issue.

  1. The chi-square magnitudes seem roughly proportional to the bin count total for a given channel. That is, the larger the peak of the decay curve, the larger the chi-square value. Because of this fact, Slim Plotter currently divides the chi-square value for a given channel by that channel's peak value.
  2. I believe the chi-square values returned by Slim Plotter's  LM algorithm are raw values, not divided by degrees of freedom. So Slim Plotter also divides chi-square by degrees of freedom (num - params.length, which I am not entirely certain is correct).

My knowledge of statistics is insufficient to judge whether this approach is correct. Further research is warranted to figure out the best way to report useful chi-square values, or other values indicative of the fit's relative "goodness."

Change History

comment:1 Changed 13 years ago by curtis

  • Description modified (diff)

comment:2 Changed 13 years ago by curtis

  • Status changed from new to assigned

comment:3 Changed 13 years ago by curtis

  • Owner curtis deleted
  • Priority changed from major to critical
  • Status changed from assigned to new

comment:4 Changed 13 years ago by curtis

  • Owner set to nor

comment:5 Changed 6 years ago by curtis

  • Status changed from new to closed
  • Resolution set to moved
  • Description modified (diff)
  • Severity set to serious
Note: See TracTickets for help on using tickets.