Zwavelzure-ammoniak-proeven in de Java-suikerindustrie en een methode tot bepaling van het optimum in deze proeven

Orthogonal polynomials were applied for the estimation of optimal dosage instead of, as formerly, considering only significant differences in yield of pure sugar. Polynomials of second degree fitted the observed cane yields satisfactorily. Very large observed values of the regression coefficients indicated serious disturbance; if so, the yield/rate relation was not well represented by a polynomial.

No simple mathematical function was found to represent the regression of sugar content on fertilizer rate. The correlation between cane yield and sugar percentage was a good measure for judging sugar percentage for a given rate, at least for correlation coefficients within the region -0.8 and 0. Other values pointed to serious aberrations. Large negative values of the regression coefficient of the term of the second degree and small values of that of the third degree corresponded to low correlation. If so, correlation coefficients calculated from the data, ignoring those from the lowest rate, were more useful.

If more than 30% of the cane lodged, the mathematical function was not satisfactory.