The following formula are used for linear regression. In case of geometrical regression, x and y have to be substituted by ln x and ln y. In case of exponential regression only y, and in case of logarithmic regression only x has to be substituted.
a = (Σy - b·Σx)/n b = (n·Σxy - Σx·Σy)/(n·Σx2 - Σx·Σx) Coeff. of determination r2 = h1/h2 __ Coeff. of correlation r = √r2 Variance s2 = (h2 - h1)/(n-2) __ Standard deviation s = √s2 with h1 = b·(Σxy - 1/n·Σx·Σy) and h2 = Σy2 - 1/n·Σy·Σy