Jeff,
I have a simple technique that may give you what you want. Try using the TREND GX
(X-Utility | Filters | Polynomial). Use Trend Order = 1. The "Output trend
channel" will contain a least-squares, best-fit straight line. The "Output
residual channel" will contain the differences between the original data points
and the straight line.
Just do standard Geosoft channel statistics on the "Output residual channel" to
get min, max and standard deviation of the differences. This approach has the
added advantage of doing an entire database in one step. The Linear Regression
GX will not work over multiple lines.
Gerry Connard
Jeff Gamey wrote:
> Bill-
>
> Thanks for the outline, but my application is rather more rudimentary than
> you give me credit for. I've got a simple spread of points that
> approximate a straight line. I can use the linear regression function in
> Geosoft to calculate slope and intercept, but I really want to know how
> well the points fit that line and I can't find a Geosoft function to do it.
>
> -Jeff
>
> At 02:22 PM 11/14/00 , you wrote:
> >Jeff:
> >
> >The methodology we have been using is to use the technique of bootstrapping,
> >or resampling.
> >
> >Take your original line data, randomaly select a subset of all teh data
> >points.
> >Recalculate the slope.
> >Repeat random resampling of the full data set
> >Recalculate slope.
> >
> >Repeat process a number of times...
> >
> >You can then calculate the mean slope and derive an estimate of the error
> >(standard deviation of that slope).
> >This procedure gives you a depth estimate and a error on your depth
> >estimate.
> >
> >Of course what it really serves to do is to emphasise that you really don't
> >know very much about the depth of the longer wavelength portion of your
> >spectrum since you only have a very small number of points in this portion
> >of your curve.
> >
> >Bill Morris
> >
> >PS Kristi Markham one of my grad students gave talk on this subject at last
> >years EGS conference, and is actively writing a paper on this subject.
> >
> >
> >
> >
> >----- Original Message -----
> >From: "Jeff Gamey" <[EMAIL PROTECTED]>
> >To: <[EMAIL PROTECTED]>
> >Sent: Tuesday, November 14, 2000 7:42 AM
> >Subject: [geonet]: linear regression
> >
> >
> > > I can run linear regression to get the slope and intercept of my data, but
> > > I'd like a measure of error of fit to that line. Does anyone have a way
> >to
> > > do that apart from (a) manually inputting the slope and intercept into a
> > > math expression for error of fit, then (b) calculating the std dev of the
> > > error?
> > >
> > > Having it included in the linear regression GX would be ideal. If someone
> > > wants to talk me through it, I'll try to modify it myself.
> > >
> > > *******************
> > > * Jeff Gamey (4tj)
> > > * Environmental Sciences Division [EMAIL PROTECTED]
> > > * Oak Ridge National Laboratory (865) 574-6316 tel
> > > * POBox 2008, Bldg 1505, MS6038 (865) 576-8646 fax
> > > * Bethel Valley Rd., Oak Ridge, TN (865) 599-0820 cell
> > > * 37831-6038
> > >
> > > _______________________________________________________
> > > More mailing list info http://www.geosoft.com/support/listserv/index.html
> > >
> >
> >_______________________________________________________
> >More mailing list info http://www.geosoft.com/support/listserv/index.html
>
> *******************
> * Jeff Gamey (4tj)
> * Environmental Sciences Division [EMAIL PROTECTED]
> * Oak Ridge National Laboratory (865) 574-6316 tel
> * POBox 2008, Bldg 1505, MS6038 (865) 576-8646 fax
> * Bethel Valley Rd., Oak Ridge, TN (865) 599-0820 cell
> * 37831-6038
>
> _______________________________________________________
> More mailing list info http://www.geosoft.com/support/listserv/index.html
--
*************************************************************************
Northwest Geophysical Associates, Inc. http://www.nga.com
Box 1063
Corvallis, Oregon 97339
voice: 541-757-7231 / fax 541-757-7331
_______________________________________________________
More mailing list info http://www.geosoft.com/support/listserv/index.html
