Turn on suggestions
Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.
Showing results for
Topic Options
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Mute
- Printer Friendly Page
How can I obtain the coefficient uncertainties from a linear fit?
Jason V.
Member
07-16-2004 12:17 PM
Options
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report to a Moderator
I'm doing a very simple linear fit on Current vs. Voltage data, but I can't figure out how to determine the uncertainties on the coefficients (m and b) of the fit (y = mx+b). The vi gives the MSE, which is useful, but I need the uncertainties on the coefficients. Thanks for any help offered.
07-16-2004 01:22 PM
Options
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report to a Moderator
Hmm, I think I've figured it out... just doing it by "brute force" (ie. just directly calculating the errors). If anyone knows of a vi or other method that is possibly better, I would still love to hear it.
Chris_Megdanoff
Member
01-04-2005 01:31 PM
Options
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report to a Moderator
I know this has been a while, but I happen to be dealing with a similar issue. In my case, I need to calculate the analytical uncertainty and the coefficient uncertainty is needed (linear or quadratic though). Did you find any good solutions?
shoneill
Active Participant
01-04-2005 01:55 PM
Options
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report to a Moderator
It's quite a difficult thing to calculate.
Normally when you do a regression, you're assuming that the X-values are correct and the values vary in the Y-direction.
I think in order to simulate (Calculate I think isn't really possible) the coefficient uncertatinty you need many different data sets. If you have the variance of the indiviual data points, you can maybe try making a fit on many different sets representative of the value variance and from the resulting coefficients estimate the coefficient uncertainty.
Some things are certain though (Which I know from uncertainty calculations in pH measurements)
1) You need multiple data sets, or at least an estimate of how "wrong" your data could be.
2) You mostly need some understanding how the different error sources in your measurement affect the output.
Bear in mind that the calculation of the uncertainty on a single set of data will have zero variance. Only when you have multiple data sets (which you may need to simulate) can you start even thinking about coefficient variance. But then we open a whole new can of worms...
How do you know your "multiple" data sets are representative?
Hope this helps in a little way, and good luck
Shane.
Normally when you do a regression, you're assuming that the X-values are correct and the values vary in the Y-direction.
I think in order to simulate (Calculate I think isn't really possible) the coefficient uncertatinty you need many different data sets. If you have the variance of the indiviual data points, you can maybe try making a fit on many different sets representative of the value variance and from the resulting coefficients estimate the coefficient uncertainty.
Some things are certain though (Which I know from uncertainty calculations in pH measurements)
1) You need multiple data sets, or at least an estimate of how "wrong" your data could be.
2) You mostly need some understanding how the different error sources in your measurement affect the output.
Bear in mind that the calculation of the uncertainty on a single set of data will have zero variance. Only when you have multiple data sets (which you may need to simulate) can you start even thinking about coefficient variance. But then we open a whole new can of worms...
How do you know your "multiple" data sets are representative?
Hope this helps in a little way, and good luck
Shane.
Using LV 6.1 and 8.2.1 on W2k (SP4) and WXP (SP2)
