527
Views
0
CrossRef citations to date
0
Altmetric
Articles

Locally R-optimal designs for a class of nonlinear multiple regression models

ORCID Icon & ORCID Icon
Pages 107-120 | Received 09 Jun 2022, Accepted 27 Nov 2022, Published online: 12 Dec 2022

Figures & data

Table 1. The simulation results obtained from the locally R-, D-optimal designs and the balanced design on X=[0,5] for the first-order Poisson regression model with intercept.

Figure 1. Plot of the function ϕ in (Equation3) for the R-optimal design on X=[0,5]2 for the Poisson regression model discussed in Example 3.2: (a) for β=(0,1,1) and (b) for β=(0,1,0).

Figure 1. Plot of the function ϕ in (Equation3(3) ϕ(x,β)=Q(f(x)⊤β)f(x)⊤M(ξ∗,β)−1(∑j=1pej,pej,p⊤ej,p⊤M(ξ∗,β)−1ej,p)M(ξ∗,β)−1f(x)⩽p(3) ) for the R-optimal design on X=[0,5]2 for the Poisson regression model discussed in Example 3.2: (a) for β=(0,−1,−1)⊤ and (b) for β=(0,−1,0)⊤.

Table 2. Comparison of R-optimal design with D- and A-optimal designs on X=[0,5]2 for the Poisson regression model discussed in Example 3.2.

Table 3. R-optimal designs on X=[0,3]2 for proportional hazards regression models for different β and R-efficiencies of the balanced design ξb.

Figure 2. Plots of the functions ϕ in (Equation3) for the R-optimal designs on X=[0,5]2 for the Poisson regression models discussed in Examples 4.1 and 4.2: (a) for β=(0.5,0.5) and (b) for β=(1,1).

Figure 2. Plots of the functions ϕ in (Equation3(3) ϕ(x,β)=Q(f(x)⊤β)f(x)⊤M(ξ∗,β)−1(∑j=1pej,pej,p⊤ej,p⊤M(ξ∗,β)−1ej,p)M(ξ∗,β)−1f(x)⩽p(3) ) for the R-optimal designs on X=[0,5]2 for the Poisson regression models discussed in Examples 4.1 and 4.2: (a) for β=(−0.5,0.5)⊤ and (b) for β=(1,1)⊤.

Figure 3. Plot of the function ϕ in (Equation3) for the R-optimal design on X=[0,5]2 for β=(0.5,0.5) in the Poisson regression model without intercept.

Figure 3. Plot of the function ϕ in (Equation3(3) ϕ(x,β)=Q(f(x)⊤β)f(x)⊤M(ξ∗,β)−1(∑j=1pej,pej,p⊤ej,p⊤M(ξ∗,β)−1ej,p)M(ξ∗,β)−1f(x)⩽p(3) ) for the R-optimal design on X=[0,5]2 for β=(0.5,0.5)⊤ in the Poisson regression model without intercept.

Table 4. The locally R-optimal designs on X=[0,3]3 for β=(2.5,0.5,0.5) in the proportional hazards regression models, the R-efficiencies of the balanced design ξb and the overall probability of censoring under the R-optimal design.

Table 5. The R-efficiencies of the locally R-optimal designs on X=[0,5]p1 for various misspecified β for the first-order Poisson regression models with intercept.