in which hyi stands for the average value of the yi’s (i ¼ 1; . . . ;N),
that is hyi ð1=NÞ
P
N
i¼1yi. Note that one may get the result, R2 < 0,
in the case of a very bad fit using a nonlinear regression [20,21].
If a and b in Eq. (11) are adjustable parameters, then R2 coincides
with Pearson’s correlation coefficient r2 for linear regression
(R2 ¼ r2). On the other hand, if a or b has a fixed value (e.g., b ¼ 0)
or if the model is non-linear, this is no longer true, but R2 may be
used to assess the goodness of fit of the data by the function ^yðtÞ
[19,20].
In what follows we will suppose that qe is known from experiment,
which is generally the case.