MATH 376 -- Probability and Statistics 2
April 5, 2006
Example 1.
Normal equations for fitting the linear model
to the data points [0,0,2], [1,0,1], [0,1,4], [1,1,5]
> | X:=matrix([[1.0,0,0],[1,1,0],[1,0,1],[1,1,1]]); |
> | Z:=matrix([[2],[1],[4],[5]]); |
> | XtX:=multiply(transpose(X),X); |
> | XtZ:=multiply(transpose(X),Z); |
> | beta:=linsolve(XtX,XtZ); |
> | PlP:=plot3d(beta[1,1]+beta[2,1]*x+beta[3,1]*y,x=-0.5..1.5,y=-0.5..1.5,axes=boxed): |
> | SP1:=plot3d([0.05*cos(theta)*sin(phi),0.05*sin(theta)*sin(phi),2+0.05*cos(phi)],theta=0..2*Pi,phi=0..Pi,color=blue): |
> | SP2:=plot3d([1+0.05*cos(theta)*sin(phi),0.05*sin(theta)*sin(phi),1+0.05*cos(phi)],theta=0..2*Pi,phi=0..Pi,color=blue): |
> | SP3:=plot3d([0.05*cos(theta)*sin(phi),1+0.05*sin(theta)*sin(phi),4+0.05*cos(phi)],theta=0..2*Pi,phi=0..Pi,color=blue): |
> | SP4:=plot3d([1+0.05*cos(theta)*sin(phi),1+0.05*sin(theta)*sin(phi),5+0.05*cos(phi)],theta=0..2*Pi,phi=0..Pi,color=blue): |
> | display3d(SP1,SP2,SP3,SP4,PlP,scaling=constrained); |
We see the plane in
that comes closest to containing the 4 data points.
Example 2.
Normal equations for fitting the linear model
Y =
to the data points [0,1], [1,4], [2,7], [3,8].
> | with(linalg): |
We let
and
The normal equations are
> | X:=matrix([[1.0,0,0],[1,1,1],[1,2,4],[1,3,9]]); |
> | Y:=matrix([[1],[4],[7],[8]]); |
> | XtX:=multiply(transpose(X),X); |
> | XtY:=multiply(transpose(X),Y); |
> | beta:=linsolve(XtX,XtY); |
> | with(plots): |
> | Pts:=[[0,1],[1,4],[2,7],[3,8]]; |
> | PP:=plot(Pts,style=point,symbol=circle,color=blue): |
> | RP:=plot(beta[1,1] + beta[2,1]*x + beta[3,1]*x^2,x=-1..4): |
> | display(PP,RP); |
> |