CURFIT.BA

          NON-LINEAR REGRESSION ANALYSIS PROGRAM

             (C) COPYRIGHT 1987 David O. Rowell
                 All rights reserved
           3961 Cherry Valley Turnpike, Marietta, NY  13110 - CIS 70406,445

          Permission is granted to members of CompuServe's Radio Shack MOD
          100 SIG for their personal use of this program.  Sale,distribution
          or any other use of the program is forbidden without the written
          permission of the copyright holder.  This copyright notice must
          be retained in all copies of the program.

          Regression analysis is a method used to determine statistical
          relations between variables.  By far the most utilized
          relationship is the linear function, but many others are commonly
          useful.  Regression analysis deals with paired observations of
          data, determining the constants of the straight line of best fit
          to that data.  In a non-linear analysis the data is first
          transformed into a linear model and then analysed.  If X is the
          independent variable and Y the dependent variable then this
          program will determine the coefficients a and b of the equation
          Y = a + bX for the transformed data.  Six curve types are
          sequentially fit to the input data: Y=a+bX, Y=ae^(bX), Y=aX^b,
          Y=a+b/X, Y=1/(a+bX) and Y=X/(a+bX).  The program also determines
          the correlation coefficient and F-ratio of the analysis.  These
          are measures of how well the regression equation fits the data.
          In general, the closer the correlation coefficient is to 1 the
          better the fit.  The higher the F ratio the better the fit.  The
          F ratio should be compared to published tables to determine a
          more quantitative measure.

          The program uses a brute force approach to the calculations,
          relying on the double precision operations of the Model 100 to
          retain accuracy in the analysis.

          The title screen serves as a seperator between the consecutive
          display of results for each of the six curve types.  The first
          display screen shows the curve type, the values for the
          coeficients "a" and "b", the correlation coeficient and F ratio.
          If you want to skip the rest of the analysis for that curve type
          press "ESC" to go on to the next curve type.  The next display is
          the analysis of variance table. Subsequent screens display the
          data alongside the calculated Y value and the percent difference
          from the data.  You are then given the opportunity to calculate
          and display Y values for keyboard entered X's.  This sequence
          repeats for each of the six curve types.  Don't forget that the
          "PRINT" key will send any screen display to your printer for a
          record.




          Your data is entered as data statements.  The first datum MUST be
          the number of data points (pairs) that will be entered.  The data
          points are then entered in additional data statements.  Those
          data must be entered as X1,Y1,X2,Y2 ... Xn,Yn.  You may use line
          numbers from 100 to 900 for your data.  The example below gives the
          DATA statements for 14 observations whose independent variable,
          dependent variable data pairs are contained in the remaining 
          statements.  The data given could be used to demonstrate the
          operation of the program by simply removing the line number and
          tic mark leaving DATA statements 100 to 120 as real data.

          	EXAMPLE DATA
          	100 DATA 14
          	105 DATA 10,21.2, 10,19.9, 11,22.5, 11,23.7, 12,25
          	110 DATA 15,30.3, 17,36.1, 19,38.6, 20,41.5, 20,42.7
          	115 DATA 23,45, 25,50, 27,53.9, 30,62.1