J. Siska et Cr. Hurburgh, PREDICTION OF WISCONSIN TESTER BREAKAGE SUSCEPTIBILITY OF CORN FROM BULK-DENSITY AND NIRS MEASUREMENTS OF COMPOSITION, Transactions of the ASAE, 37(5), 1994, pp. 1577-1582
An equation was developed to predict corn breakage susceptibility base
d on the protein content, oil content, starch content, kernel density,
and test weight. Reference values of breakage susceptibility were mea
sured by Wisconsin Breakage Tester. Two statistical techniques were us
ed to design the prediction equation, multiple linear regression (MLR)
and principal factor method (PFM). The calibration sample set include
d 94 marker corn samples. This set was used to predict corn breakage s
usceptibility from protein content, starch content rest weight, and ke
rnel density with MLR. Two sample sets were used for PFM calculations.
One sample set of 252 corn samples was used to calculate standardized
scoring coefficients of a common factor, which represented protein, o
il, starch, and density. The other set, the same calibration sample se
t from MLR analysis, was used to regress breakage susceptibility again
st the common factor and test weight. The calibrations were validated
on 36 samples not included in any of the sample sets described For the
MLR analysis, the R(2) (validation) was 65%, and the standard error o
f prediction (SEP) was 1.0 percentage point. The PFM equation gave R(2
) of 64% and the SEP was 1.0% point.