This study by Professor John Kalivas of Idaho State University is supported by the Analytical and Surface Chemistry Program of the Chemistry Division and the Statistics Program of the Division of Mathematical Sciences under the umbrella of the NSF-wide Mathematical Sciences Priority Area. This is a statistical study aimed at improving analysis of spectroscopic data. These activities fall in the specialized field called "chemometrics." The idea is to develop multivariate calibration, which is of interest when the number of predictors exceeds the number of samples (p>n, rather than the usual assumption n>>p). Professor Kalivas proposes three new model approaches that address the general problem of the transfer of a calibration model developed on one instrument to other instruments, or the problem of building a calibration model for several instruments.
All three models are based on Tikhonov Regularization. Each uses both variance and bias to address the two issues of harmony and harmony/parsimony balance. The work is being done by undergraduates using the program MATLAB.
Efficiency and effectiveness of data analysis is becoming increasingly important in all of the quantitative sciences, including biology. The integration of state of the art statistical methods within the natural science disciplines is required for advancement in these fields. Education of undergraduates in these advanced methods prepare them for any type of scientific professional pursuit.
