TABLE 2.

Main steps in population model building for oral data^a

Stepb	Model	Objective functionc	Commentary
1	Monocompartmental CL, V, ka	1,016	ς2 = 7.61
2	linearly related to PNA	666	ς2 = 0.470; PNA influences
3	linearly related to PNA, linearly related to BW	648	ς2 = 0.291; BW influences
5	linearly related to PNA, proportional to BW	650	ς2 = 0.288; is proportional to BW
8	= θ1 · GFR · (BSA/1.73), GFR = (7.2 × PCAθ3)/(θ2θ3 + PCAθ3)	560	ς2 = 0.219; this clearance model is better
10	= θ1 · GFR (BSA/1.73) · (40/SCR)	583	ς2 = 0.202; scatterplots are better than those in step 8
16	Similar to step 10, but with zero-order absorption rate	596	ς2 = 0.554; the fit is worse than that in step 10
17	Bicompartmental model: = θ1 · GFR (BSA/1.73) · (40/SCR) = θ2 · BW = θ3 = θ4 · BW = θ5 · BW	577	ς2 = 0.183; not significantly better than the one-compartment model

^aAbbreviations: θ₁ to θ₆, population parameters to be estimated; ς², common error variance. To simplify notations, the subscript j on typical parameter values and covariates, the normalization of the covariates with respect to their median, and the ratio with F have been omitted. After step 5, has the same definition in all remaining steps. After step 8, GFR has the same definition in all remaining steps. 

^bSome steps have been omitted for brevity. 

^cFunction to be minimized. The critical change is 3.84 for 1 degree of freedom at the 0.05 level.