AnsweredAssumed Answered

vrf definition of goodness of fit ?

Question asked by VRFuser on Jun 10, 2001
Hi Richard,
           Looked like a late night for you...

>From what I remember of my stats at school, the correlation needed to be up
around 0.9 to 0.95 to be able to make any meaningful conclusions so 1 would
be a perfect fit.  I don't have my A level notes to hand, unfortunately,
but from "Quantitative Methods" by Diana Bedward it is quoted as:

( sum(x*y) - n*mean_x*mean_y )/root((sum x^2 - n*mean_x^2)*(sum y^2) -
n*mean_y^2))

which follows from the least-squares method of doing the regression; if the
'means' coincide with the 'actuals' then the top and bottom of the equation
come out to be numerically the same.  Note that a negative slope yields a
perfect correlation figure of -1.


Hope that helps,

Regards,

Mike Watts
---------------------------------------------------------------------
This is the "vrf" maillist, managed by Majordomo.  To send messages to
this maillist, just email to "vrf@lvld.agilent.com".  Subscriptions and
unsubscriptions are done through the address "vrf-request@lvld.agilent.com".
If you need details, just send a message containing the text "help"
to "vrf-request@lvld.agilent.com".
---------------------------------------------------------------------

Outcomes