Opened 12 years ago

Closed 12 years ago

#305 closed bug (fixed)

Large standard deviation values from linear fits?

Reported by: cwelch@… Owned by: srkline
Priority: major Milestone: Release 7.0 Maintenance
Component: Analysis Keywords:
Cc: Blocking:
Task:

Description

When I use the linear fits package in v. 7.03, I get very large standard deviation values. For example, when doing a Guinier plot to obtain Rg, I might get Rg = 44.142 +/- 5880 Angstroms. For this same example, the Chi-Squared value was 8.8941e-09. With previous versions of the NCNR Igor macros, I did not encounter such problems. Is this a known issue, or have I inadvertently set some parameter to the wrong value or done something else wrong? I do not encounter similar problems when using the Curve Fit features.

Change History (1)

comment:1 in reply to: ↑ description Changed 12 years ago by srkline

  • Milestone set to Release 7.0 Maintenance
  • Resolution set to fixed
  • Status changed from new to closed

The /I=1 flag should be I=0, since the weights are calculated as 1/err -- since this procedure was written way back in time when that was the only acceptable way for the weights to be cast. fixed in r746

Note: See TracTickets for help on using tickets.