The authoress polemises with H. Dudek's paper (Ibid. 2003, no 2 pp. 41-51) and claims that the procedure applied is improper. From the theorems proved in the present paper it follows that the values of the pairwise correlation coefficients of explanatory variables with perfect multicollonearity may be analysed independently on both the variances and the parameters of linear functions concerning the relations between these variables. It was shown that perfect multicollinearity of explanatory variables in the econometric models may occur even though all pairwise correlation coefficients of these variables are very small in absolute values, e.g. when all pairwise correlation coefficients of 'k' explanatory variables are equal to ' -1/(k-1)' then these variables are perfectly multicollinear. This raises the question whether it is sensible to use those methods of explanatory variable selection in which low pairwise correlation of all explanatory variables is treated as one of the necessary conditions for the good quality explanatory variable set.
JavaScript is turned off in your web browser. Turn it on to take full advantage of this site, then refresh the page.