Yes, it is possible for the values of an NIR calibration model to be lower than the prediction. This can happen due to various factors, including: * Model Overfitting: If the calibration model is overfitted to the training data, it may not generalize well to new, unseen data. This can lead to predictions that are higher than the actual values. * Measurement Errors: Inaccurate measurements during the calibration process can introduce bias into the model, leading to underestimation or overestimation of predictions. * Sample Heterogeneity: If the samples used for calibration are not representative of the population being studied, the model may not accurately predict the values of new samples. * Data Preprocessing: Incorrect or inappropriate data preprocessing techniques can also affect the accuracy of the calibration model. To address this issue: * Careful Model Development: Ensure that the calibration model is developed using appropriate techniques and that it is not overfitted to the training data. * Rigorous Data Quality Control: Implement strict quality control measures to minimize measurement errors and ensure data accuracy. * Representative Sampling: Use a representative sample of the population being studied to develop the calibration model. * Appropriate Data Preprocessing: Use appropriate data preprocessing techniques to remove noise and improve model performance. By addressing these factors, it is possible to develop more accurate NIR calibration models that provide reliable predictions.

Article Optimisation of the predictive ability of NIR models to esti...