Following changes were made:
- The covariance matrix is now calculated independent of chi squared
- The parameter uncertainties (parerr) are calculated as the square root out of the covariance matrix times chi squared divided by m-n
- For weighted data, the parameter uncertainties have to be taken out of the square root of the covariance matrix instead of parerr
- lmcurve2 was adapted to lmmin2 with covariance matrix and returns in parerr the weighted parameter uncertainties
- the curve2 demo was adapted to the changes in lmcurve2
- the man pages were updated to explain these changes
Calculated parameter uncertainties for unweighted case were compared with levmar and gsl. Calculated parameter uncertainties for weighted case (curve2 with lmcurve2) were compared with gsl. For both comparisons the calculated parameter uncertainties matched those of levmar and gsl.