The unique solution to the normal equations exists only if the inverse ofX'Xexists. This, in turn, requires that the matrix X be of full columnrank; that is, there can be no linear dependencies among the independent variables. The practical implication is that there can be no redundancies in the information contained in X. For example, the amount of nitrogen in a diet is sometimes converted to the amount of protein by multiplication by a constant. Because the same information is reported two ways, a linear dependency occurs if both are included inX. Suppose the independent variables in a genetics problem include three variables reporting the observed sample frequencies of three possible alleles (for a particular locus). These three variables, and the 1 vector, create a linear dependency since the sum of the three variables, the sum of the allelic frequencies, must be 1.0. Only two of the allelic frequencies need be reported; the third is redundant since it can be computed from the first two and the column of ones