KIMBALL
were determined. The equation used for these models was (6) P=K,V+K,li, (I?
where P = transpulmonary pressure, V = tidal volume, ri = airflow, K1 = reciprocal of dynamic compliance, and K2 = inelastic airway resistance.
It should be pointed out that the "constants" K1 and K2 are really not constants.
In a biological system, these numbers are in fact random variables, and the number assigned to a particular variable is the estimate of the mean of this variable. Consequently, the assigning of a number to these constants should be done on a proper statistical basis. For example, one procedure (8,18) which is often used to determine K, is to divide P by I/ when ri = 0. Not only is the procedure statistically incorrect, but it suffers from the usual problems associated with the determination of accurate zero crossings in the presence of noise. However, much work was performed on this simple model by these early investigators using analog computers (7, 22, 13, 26). The use of analog computation had the advantage that it was easy, inexpensive, and provided an on-line parameter determination.
Later, the gradual replacement of analog computers by digital computers provided a much more powerful tool for the modeling of biological systems (IZ). As a result, the nonlinear aspects of respiratory mechanics were modeled, and the number flood appeared.
Since this study was associated with a clinical unit where the numbers from a respiratory model were to be used to determine therapeutic maneuvers, it was of importance to use the simplest possible model on both a physiological and computational basis. Therefore, it was important to determine whether actual patient data contained sufficient information to warrant higher order models. This question was answered by a comparison of least-squares analysis on a set of proposed models (6) which were of the form