253
til
where Yi is the value of 1 measurement and Yi is the
obs calc
estimated 1th measurement. The above method is valid only when all
observations in Y have the same, though unknown, variance.^ When the
observed values do not have the same variance, each point should be
weighted by the inverse of its variance.*^ If it can be assumed that
the variance in Y is proportional to Y, then appropriate weighting
factor is 1/Y2.
Fitting of plasma data of buprenorphine. In pharmacokinetics,
proper estimation of the parameters of the plasma concentration-time
profile is dependent upon:
a) Compartment model chosen.
b) Number of experimental points, the time interval between these values
and the number of half-lives over which the samples were collected.
c) Weighing of data.
d) Analytical sensitivity, especially at low plasma concentrations.
The classical method of estimating the parameters of an equation
describing the plasma concentration-time profile of a drug conforming to
2 or 3-compartment body model is by "stripping", the method of
53i
residuals. In this technique, least square estimates of the terminal
phase are calculated. For a 2-compartment body model, the parameters of
the initial phase are estimated from a plot of the difference between
the initial values and the extrapolated terminal phase.~^1
As an example, for dog C at 1.439 mg/kg IV bolus dose of 1 (Table
2), fitting of the plasma data by the method of residuals is given in
Fig. A-ld. The sum of squares calculated in accordance with equation A3
is 81170. These initial estimates of the parameters are given in Table
Al. These parameters were used as initial estimates for the computer