|Lupton, C - TX A&M U, SAN ANGELO|
|Pfeiffer, F - TX A&M U, SAN ANGELO|
|Minikhiem, D - TX A&M U, SAN ANGELO|
Submitted to: Journal of Animal Science Supplement
Publication Type: Abstract Only
Publication Acceptance Date: June 21, 1995
Publication Date: N/A
Technical Abstract: The objective was to determine the expected level of precision for prediction of mohair quality using near-infrared spectroscopy (NIRS). Fifty-nine mohair fleeces, collected at the Texas A&M Research Station, San Angelo, were sampled four times each. Each sample was subjected to monochromatic light from a Pacific Scientific near infrared reflectance spectrophotometer fitted with a transport mechanism. This mechanism allowed the instrument to scan a 54 cm2 segment of the fleece sample. Thirty-two scans were taken and averaged for each of the four samples of each fleece. The samples were then subjected to conventional procedures to determine mohair yield, vegetable matter content, fiber diameter, and percent med and kemp fibers. The data were used to develop prediction equations which related NIRS spectra to laboratory determined yield and diameter parameters. The same samples were predicted with the equations and the results, along with the conventional laboratory values, were subjected to a simple analysis of variance to determine variation within and across samples. Coefficient of variation for mohair base and diameter were 1.0 and 1.4, whereas those for predicted mohair base and diameter were 1.4 and 3.4, respectively. The increased variation within samples for predicted values represents sampling error and lack of fit between NIRS and the laboratory determined values. Sampling error could be large since only the surface was scanned, albeit the surface was much larger than that used for routine forage analysis. The entire sample was used for conventional laboratory analyses. Calibration and cross validation statistics are presented in the table below for each constituent.