Making the best use of sensor data requires close attention to calibration and data analysis. To this end, our group researches various linear, nonlinear, and chemometric approaches to these issues. For instance, best linear unbiased estimation provides a method by which the resolution of a sensing system may be predicted, and which under appropriate conditions will provide measurements with optimal resolution. To apply this technique the system response (i.e. derivatives) and measurement statistics must be known.

Then the optimal estimate x* and covariance matrix of a set of system parameters determined from measurements y* is given by

These equations may be used to characterize the effectiveness of sensing systems and data analysis algorithms.

This shows the uncertainty in two-color SPR measurements, predicted by best linear unbiased estimation.

This shows the resolution attained using centroid and quadratic fit algorithms, predicted by best linear unbiased estimation.