 |
A fallacy in
regression
Finds the parameters of a certain function or its logarithmised
form (with varied results). |
|
|
| n, nv, m, itransform |
0
1
|
No. of points, no. of independent variables, no. of parameters,
"transform" data. • |
| X |
|
X-values given in "nv" rows. • |
| Y |
|
Y-values. • |
| krit |
1
2 (min. sq.)
3
4
∞ (minmax) |
Criterion [power of (ycalc−y)]. |
| tol, maxcal |
|
Tolerance, max. no. of iterations. • |
| P |
|
Parameters (of order unit). • |
| Z |
|
Scale factors. True values are P.Z. |
Solves a regression for the function below (left-hand
equation) or its transformation into a logarithmised, linearised form
(right-hand equation). The values of K and a differ, mainly
for the former (about 10 % with the data proposed). (The values would be
identical if the data were "perfect".) The original form is
the legitimate one, although the linearised version is commonly preferred.
|
|
|
|
