Evaluation location: , United States
Samples reported upon are from the USDA Germplasm Collection based in Urbana, IL. Soybean samples were scanned by NIR at the University of Minnesota's Soybean Breeding Project laboratory. Whole soybean samples received from USDA-NASS were first ground and then scanned on a FOSS 6500 NIR Instrument. NIR Spectra from the FOSS 6500 were predicted using ISIPredict Software version 1.10.2.4842. Calibrations, provided by FOSS North America, were used to predict soybean composition from the NIR spectra. Presented as a comment for each reported compositional value is a value for Global "H". As stated in FOSS Application Note FNA538-01. "Global H value-measures how far this sample is from the center of the calibration database (The samples selected for this spectra). The distance is normalized and a value exceeding 3 is typically considered an outlier." A number of factors can contribute to a sample being flagged as an outlier when using a given set of calibration equations. With soybeans, this could include seed coat color if the calibrations were developed using only soybeans with yellow seed coats. In addition to soybeans with yellow seed coats, this dataset includes soybeans bearing other colors of seed coat. Sample preparation and presentation to the instrument can also contribute to higher H values. In addition to the previously stated reasons, another reason for a high H value could be that a sample is truly unique.