The testing and adjusting procedure of near infrared (NIR) spectrophotometers is based on the measurement of some standards and, if necessary, on the adjustment of the constants in the calibration equation. For this work some use few standards, whereas others use 20 or more. This work was aimed to determine the range of error compensation and the minimum number of standards required. The experiments were applied to wheat protein measurement using two scanning spectrophotometers. The errors in the NIR measurements were characterised as bias, skew, error derived from skew ( Eskew) and standard error of difference corrected for bias and skew ( SEDc) parameters and supposed that errors derived from the change in the wavelength or reflectance of the instrument. The confidence intervals of bias and skew, derived from duplicate measurements of various numbers of wheat standards, were used to determine the minimum number of standards required. The range of error compensation was defined with those bias values, where SEDc was smaller, than an acceptable limit. The range of compensation corresponded to a bias value of ± 8 g kg−1 for wheat protein measurements. The detection of error of measurements required 4 wheat standards. The elimination of errors of bias and skew required 9 standards within the above limits. The developed procedure was tested in case of real instrument error. Diminishing a bias from 5.2 g kg−1 to 0.7 g kg−1 and the root mean square difference ( RMSD) to an acceptable level required the use of 9 standards, similar to the model experiment. The simplicity and rapidity (about 10 min) of the procedure enabled the routine test of NIR instruments. The range of error compensation and the number of standards referred to wheat protein. The simple modelling procedure proved also suitable for the determination of these values for other components and under other measuring conditions.
Detection of chalk in single kernels of long‐grain milled rice using imaging and visible/near‐infrared instruments
