Least square fitting with one parameter less

It is shown that whenever the multiplicative normalization of a fitting function is not known, least square fitting by ${\chi }^2$ minimization can be performed with one parameter less than usual by converting the normalization parameter into a function of the remaining parameters and the data.