Attempts to quantitatively analyze time series of blood hormone levels present a number of rather novel and difficult issues that be addressed during analysis.

Two particularly analytically challenging properties often displayed by hormone concentration time series data are their episodic pulsatility and their temporal pattern irregularity. More "conventional" time series analysis methods generally fail quite miserably at providing quantitative information useful for the interpretation of such data.

Progress in understanding both normal and pathophysiologic endocrine secretory functioning is contingent upon reliable quantitation of such hormone concentration profiles.

In close collaboration between basic and clinical research endocrinologists and quantitative scientists and software developers, a number of algorithms have been, and continue to be, developed for pulse detection, deconvolution for estimating secretory rate profiles, event coincidence analysis, and assessment of temporal pattern irregularity. These algorithms devote considerable attention to confidence interval estimation on the derived results, by both conventional statistical methods and, to an increasing extent, by means of empirical resampling strategies.

Faculty associated with this research include:

William S. Evans

Michael L. Johnson

Michael O. Thorner