#277 SciDAVis returns "nan" for parameters of the fit if a point has an error equals to zero

[Fellype]

SciDAVis returns "nan" for parameters of the fit (parameters, errors, chi^2, R^2,...) if a point has an error equals to zero or if it is not specified (empty cell).
To reproduce the bug:
In a table with columns X,Y, and YErr:
1) change the value of one YErr cell to zero
2) create a plot with error bars
3) try to fit the curve with any fit function
-> look at the results log and you will see the "nans" for all fit parameters

Tested with versions 1.D8 and 1.D9 on Slackware Linux.

[Fellype]

Actually it is not a SciDAVis bug, but it is related with the way that parameters are calculated using the Least Squares method when the data have y-errors [1]. In this case, an error equals to zero leads to an infinity weight, which results in the "nans".

[1] https://en.wikipedia.org/wiki/Least_squares#Weighted_least_squares