MULTIVARIATE DATA MAY INCLUDE MISSING VALUES FOR VARIOUS REASONS. MISSING VALUES CAN BE ESTIMATED USING DIFFERENT APPROACHED, SUCH AS INTERPOLATION, NIPALS LIKE PROCEDURE AND ETC [1].A MAIN GOAL IN THE Calibration METHODS SUCH AS CLS, ILS, PCR AND PLS IS THE PREDICTION OF UNKNOWN SAMPLES USING THE ESTIMATED MODEL [2]. IN THIS STUDY MULTIVARIATE Calibration PURPOSES IS ACHIEVED BY USING MISSING VALUES ESTIMATION STRATEGIES. CONSEQUENTLY, MULTIVARIATE PREDICTION OF UNKNOWN SAMPLES IS POSSIBLE WITH NO NEED TO CONSTRUCTION AND APPLICATION OF Calibration MODELS.STANDARD ERROR OF PREDICTION (SEP) SET AS THE CRITERION FOR EVALUATION OF THE PROPOSED METHOD. FOR THIS PURPOSE NIPALS, KIERE, WOLD, AND ITERATIVE IMPUTATION METHODS WERE EXAMINED. FINALLY THE SEP OF Calibration METHODS WERE COMPARED WITH THE SEPS FROM THE PROPOSED METHOD. AMONG THE Calibration MODELS, PLS RESULTED IN LOWEST SEP VALUES.THE METHODS OF HANDLING MISSING VALUES WERE SUCCESSFULLY USED AS FOR Calibration TRANSFER PURPOSES [3], AS WELL. Calibration TRANSFER IS UTILIZATION OF Calibration MODEL FROM ONE CONDITION TO ANALYSIS THE DATA OF A SECOND CONDITION. TRANSFER OF Calibration FROM ONE INSTRUMENT TO ANOTHER IS AN EXAMPLE. A NOVEL POINT IN THIS APPROACH IS THE POSSIBILITY OF Calibration TRANSFER IN THE ABSENCE OF TRANSFER SAMPLES, WHICH IS AN ADVANTAGE TO COMMON Calibration TRANSFER TECHNIQUES.

A COMMON SOLUTION TO PROBLEM OF RANK DEFICIENCIES AND COMPLETE RESOLVING OF IT IS COLUMN AUGMENTATION [1]. IF THERE IS NOT ANY DIFFERENCE BETWEEN PURE SPECTRA OF COMPONENTS IN DIFFERENT DATA MATRICES, THEN ANALYSIS OF AUGMENTED DATA GIVES THE PARAMETERS AND CORRECT SPECTRAL PROFILES.HOWEVER IN THE CASE OF INDIVIDUAL DATA ANALYSIS, ONLY THE PARAMETER CAN BE ESTIMATED CORRECTLY AND NOT SPECTRAL PROFILES [2-3]. IN PRESENCE OF SPECTRAL DIFFERENCES IT IS NOT POSSIBLE TO AUGMENT THE DATA.Calibration TRANSFER MAKES IT POSSIBLE TO AUGMENT THE DATA. IN THIS STUDY, FIRST, Calibration TRANSFER METHOD WAS USED AS AN ADDITIONAL STEP IN A HARD MODELING PROCEDURE TO CORRECT THE SPECTRAL DIFFERENCES BETWEEN THE TWO DATA MATRICES [3]. HOWEVER IN FIRST STEP, THE RANK DEFICIENT EQUILIBRIUM SYSTEM CANNOT BE RESOLVED COMPLETELY AND ONLY THE PARAMETERS CAN BE ESTIMATED CORRECTLY AND LINEAR COMBINATIONS OF PURE SPECTRAL PROFILES ARE OBTAINED.IN THE NEXT STEP, AFTER CALCULATION OF CORRECT PARAMETERS, A NEW METHOD WAS PROPOSED TO CALCULATE THE PURE SPECTRAL PROFILES OF THE CONSIDERED RANK DEFICIENT SYSTEM. THE PROPOSED NEW METHOD IS BASED ON PIECE WISE DIRECT STANDARDIZATION (PDS) METHOD. HOWEVER, THIS STEP IS ANOTHER NEW APPLICATION OF Calibration TRANSFER IN CALCULATION OF SPECTRAL PROFILES. THIS METHOD GIVES A RANGE OF ACCEPTABLE PURE SPECTRAL PROFILES AND NOT UNIQUE SPECTRAL PROFILES. IN ORDER TO INVESTIGATE THE POTENTIAL OF PROPOSED METHOD, THREE DIFFERENT SIMULATED EQUILIBRIUM SYSTEMS WERE USED.