Date: Thu, 29 Oct 1998 14:53:11 GMT
Reply-To: Roland Leong <roland@SHOTGUNREPORT.COM>
Sender: "SPSSX(r) Discussion" <SPSSX-L@UGA.CC.UGA.EDU>
From: Roland Leong <roland@SHOTGUNREPORT.COM>
Organization: Auric, Inc.
Subject: Curve estimation
Group:
I'm a newbie; please forgive my ignorance.
I have SPSS Student Version 6.1.3 and the SPSS 6.1 Guide to Data Analysis,
which I've read.
The book does not cover the use of
Statistics->
Regression->
Curve Estimation
at all.
Is this menu item where you can test how well the data fits the different
types of functions. e.g. quadratic and logarithmic, without having to
transform the data?
I did this with a small set of data, checking off linear and quadratic,
and the output was:
========================
Dependent Mth Rsq d.f. F Sigf b0 b1 b2
CONTRIBU LIN .953 8 162.46 .000 7931775 12.1842
CONTRIBU QUA .955 7 74.24 .000 7055649 13.7970 -5.E-07
========================
Because the R square was slightly higher and the Sigf was small, I guessed
that the quadratic would be a better fit.
I then transformed my data by adding a variable that was the square of my
independent variable and did a stepwise regression with the independent
variable and the square of the independent variable.
=======================
Multiple R .97625
R Square .95307
Adjusted R Square .94720
Standard Error 2735320.9801
Analysis of Variance
DF Sum of Squares Mean Square
Regression 1 1215541351209185.00 1215541351209185
Residual 8 59855846914987.6000 7481980864373.45
F = 162.46250 Signif F = .0000
---------------------- Variables in the Equation -----------------------
Variable B SE B 95% Confdnce Intrvl B Beta
FUNDRAIS 12.184219 .955919 9.979868 14.388570 .976252
(Constant) 7931774.9020 1465976.954 4551229.4496 11312320.354
----------- in ------------
Variable T Sig T
FUNDRAIS 12.746 .0000
(Constant) 5.411 .0006
------------- Variables not in the Equation -------------
Variable Beta In Partial Min Toler T Sig T
FUNDSQR -.136406 -.201652 .102565 -.545 .6029
========================
The answer used only the independent variable and knocked out the
independent variable squared.
I also ran a linear regression with the independent variable squared
===========================
Multiple R .91084
R Square .82964
Adjusted R Square .80834
Standard Error 5211561.9663
Analysis of Variance
DF Sum of Squares Mean Square
Regression 1 1058114173099310.00 1058114173099310
Residual 8 217283025024862.100 27160378128107.8
F = 38.95801 Signif F = .0002
---------------------- Variables in the Equation -----------------------
Variable B SE B 95% Confdnce Intrvl B Beta
FUNDSQR 3.18204E-06 5.0981E-07 2.00642E-06 4.35767E-06 .910843
(Constant) 15534198.625 2038049.091 10834453.812 20233943.438
----------- in ------------
Variable T Sig T
FUNDSQR 6.242 .0002
(Constant) 7.622 .0001
End Block Number 1 POUT = .100 Limits reached.
===========================
QUESTION:
Why does the Curve estimation give the indication that a quadratic formula
would be the choice, but when I run a quadratic regression it says no go.
