Regression

 BBS: Inland Empire Archive
Date: 04-11-92 (17:14)             Number: 11
From: JEAN CREPEAU                 Refer#: NONE
  To: DON KROUSE                    Recvd: NO  
Subj: Regression                     Conf: (2) Quik_Bas

In a message to ALL, DON KROUSE wrote:
DK=> Hi Folks,

DK=>  Does  anyone  know  were  I  might  find  source  code for a multiple
    regression routine?

DK=>  I  am  particularly  interested  in  a routine that does Linear, Log,
    Exponential, Power,  etc. and  that optimizes  the order  in which  the
    dependent variables are included.

Use the following formula

        m=(n*SUM(x*y) - SUM(x)*(SUM(y))/(n*SUM(x*x) - SUM(x)*SUM(x))
        b=(SUM(y)-m*SUM(x))/n
                where n=# of samples

        Your linear equation is y=mx+b

Here is the source:

TYPE RECT
X AS SINGLE
Y AS SINGLE
END TYPE

SUB REGR(N,VEC() AS RECT,MODE,M,B)
FOR X=0 TO N-1
        IF MODE AND 1 THEN CX=LOG(VEC(X).X) ELSE CX=VEC(X).X
        IF MODE AND 2 THEN CY=LOG(VEC(X).Y) ELSE CY=VEC(X).Y
        SX=SX+CX:SY=SY+CY:SXY=SXY+CX*CY:SX2=SX2+CX*CX
NEXT
M=(N*SXY-SX*SY)/(N*SX2-SX*SX)
B=(SY-M*SX)/N
END SUB

        This sub calculates m and b from a list of points vec(0...n-1).  It
has been modified to compute linear, log, exp and power regression. To  use
it:
1. You put the data in a vector of type RECT
        INPUT "NUMBER OF DATA: ",N
        REDIM VEC(N-1) AS RECT
        FOR X=0 TO N-1
                INPUT "(X,Y): ",VEC(X).X,VEC(X).Y
        NEXT
2. You determine the mode
        MODE=0 => linear
        MODE=1 => log
        MODE=2 => exp
        MODE=3 => power
3. You call the sub
        REGR N,VEC(),MODE,M,B
4. Your regression equation is
        Mode 0: y=m*x+b
        Mode 1: y=m*log x + b
        Mode 2: y=exp(b)*exp(m*x)
        Mode 3: y=exp(b)*x^m
---
 * Origin: INTERACESS Montreal (QC) Canada (514) 528-1415 (1:167/280)