Model functions

leg() function

leg(v,[-]n)
forms n+1 legendre polynomials of order 0 (intercept), 1 (linear) ... n from the values in v; the intercept polynomial is omitted if n is preceded by the negative sign. The actual values of the coefficients are written to the .res file. This is similar to the pol() function described below.

pol() function

pol(v,n) or p(v,n)
forms a set of orthogonal polynomials of order abs(n) based on the unique values in variate (or factor) v and any additional interpolated points, see !PPOINTS and !PVAL. It includes the intercept if n is positive, omits it if n is negative.

For example, pol(time,2) forms a design matrix with three columns of the orthogonal polynomial of degree 2 from the variable time. Alternatively, pol(time,-2) is a term with two columns having centred and scaled linear coefficients in the first column and centred and scaled quadratic coefficients in the second column.

The actual values of the coefficients are written to the .res file. This factor could be interacted with a design factor to fit random regression models. The leg() function differs from the pol() function in the way the quadratic and higher polynomials are calculated.

Return to start