3.2. Calibration and prediction
Different calibration models of selected segments of
wavelengths in range of 400–700 nm were developed for
predicting TSS of mango. The quality of the calibration
models was quantified by the SEC, standard error of
prediction (SEP) and the correlation coefficient R
between the predicted and measured parameter. The
optimum number of principal component (latent variables) in the case of PLS and PCR were found to be two
at the lowest level of residual validation variance and,
the lowest amount of root mean standard error of
calibration was found to be in the case of MLR
calibration for the wavelength range 440–480 nm. A
good model should have a low value forSEC, a low
value forSEP and a high value forR, but also a small
difference between the values for SEC and SEP. A large
difference indicates that too many latent variables,
principal components, are used in the case of PLS and
PCR and ‘noise’ is modelled in MLR (Lammertynet al.,
1998). The SEC, SEP andRvalues forthe MLR model
3.2. Calibration and predictionDifferent calibration models of selected segments ofwavelengths in range of 400–700 nm were developed forpredicting TSS of mango. The quality of the calibrationmodels was quantified by the SEC, standard error ofprediction (SEP) and the correlation coefficient Rbetween the predicted and measured parameter. Theoptimum number of principal component (latent variables) in the case of PLS and PCR were found to be twoat the lowest level of residual validation variance and,the lowest amount of root mean standard error ofcalibration was found to be in the case of MLRcalibration for the wavelength range 440–480 nm. Agood model should have a low value forSEC, a lowvalue forSEP and a high value forR, but also a smalldifference between the values for SEC and SEP. A largedifference indicates that too many latent variables,principal components, are used in the case of PLS andPCR and ‘noise’ is modelled in MLR (Lammertynet al.,1998). The SEC, SEP andRvalues forthe MLR model
การแปล กรุณารอสักครู่..
