Changing number of Significant Digits in Fit parameters

Share icon

Tue, 06/10/2008 - 12:23 am

Hi all. Is there a way to change the number of significant digits of the fit parameters that are displayed when I fit some data with igor. Right now, I can only see 5 significant figures which is not enough for my analysis. Any help would be greatly appreciated.

Thanks!!
The fit parameters themselves should be calculated to more significant digits than are usually displayed, but it would depend on how you are displaying them.

If you are just looking at the print out of the fit parameters and fit results that's printed to the command line after a fit, I don't know of a way to change the number of digits used there. However, doing a fit creates a wave with all of the fit parameters, and you can just inspect the values in that wave to get as many digits as you wish. Something as simple as
print/D w_coef

might do the trick for you.

June 10, 2008 at 05:58 am - Permalink

Did you by any chance also send this question to support@wavemetrics.com? It sounds just like a question I got this morning...

My answer to that e-mail was:

The coefficient wave used during the fit (usually W_coef) has the solution to full double-precision resolution. You can see that by executing this command:

print/D W_coef

If you have used a different coefficient wave, you will need to change "W_coef" to whatever name you used. If you used a different wave than the default, it is important that you used a double-precision wave.

When doing your analysis, it is best to use "W_coef[0]" or "W_coef[1]", etc., rather than copying the numbers from the history report.

It is actually unusual for a curve fit to result in a solution that is accurate to better than 5 digits. Be sure you aren't over-analyzing!


John Weeks
WaveMetrics, Inc.
support@wavemetrics.com

June 10, 2008 at 02:52 pm - Permalink

geniulo wrote:
Hi all. Is there a way to change the number of significant digits of the fit parameters that are displayed when I fit some data with igor. Right now, I can only see 5 significant figures which is not enough for my analysis. Any help would be greatly appreciated.


FYI, in general, when you have a result expressed in scientific notation, the rule regarding reporting sig figs is to never report a number past the first uncertain digit. Your uncertainty should therefore be reported (in scientific notation) to have ONE significant digit! One exception I've seen mentioned is when the least significant digit in the uncertainty is three or less. Then, reporting an additional digit in the uncertainty could be considered kosher by some.

Examples

1.456 +/- 0.478 becomes 1.5 +/- 0.5
6.856 +/- 0.378 could be 6.9 +/- 0.4 or 6.86 +/- 0.38

--
J. J. Weimer
Chemistry / Chemical & Materials Engineering, UAH

June 10, 2008 at 06:28 pm - Permalink