**Part A. Minimax Optimal Design**

- A1: G-optimal design for simple linear model with efficiency function λ(x)=x+5

Type "Goptimal" (without the double quotes) to run the program.

The model is . - A2: E-optimal designs for the Michaelis-Menten model

Type "Eoptimal" (without the double quotes) to run the program.

The model is .

- R.-B. Chen, S.-P. Chang, W. Wang, and W. K. Wong. (2011) "Optimal Experimental Designs via Particle Swarm Optimization Methods." Technique report. [download]
- Wong, W. K. (1995). On the equivalence of D and G-optimal Designs in Heteroscedastic Models. Statistics & Probability Letters, Vol. 25, 317-321. [link]
- Dette, H. and Wong, W. K. (1999). E-optimal Designs for the Michaelis-Menten Model. Statistics & Probability Letters, Vol. 44, 405-408. [link]

**Part B. Mixture Experiment**

- B1: Minimally supported D and A-optimal designs for the linear, quadratic and cubic Scheffe's mixture polynomial models with q=3 factors on the regular q-simplex:

Type "run" (without the double quotes) to run the program.

The design experimental region is the regular q simplex, . The linear, quadratic and cubic Scheffe models with q = 3 are considered. The criteria are D- and A-optimal criteria.

- W. Wang, R.-B. Chen, C.-C. Huang, and W. K. Wong. (2012) "Particle Swarm Optimization Techniques for Finding Optimal Mixture Designs." Technique report. [download]

**Part C. Locally D- and Ds-optimal Designs for Logistic Models**

- C1: Locally D-optimal designs for simple linear logistic models

Type "Doptimal" (without the double quotes) to run the program.

The logistic model is defined as . - C2: Locally Ds-optimal designs for the parameter beta in simple linear logistic models

Type "Ds_optimal_b" (without the double quotes) to run the program.

The logistic model is defined as . - C3: Locally D-optimal designs for quadratic logistic models

Type "Doptimal" (without the double quotes) to run the program.

The model is . - C4: Locally Ds-optimal designs for the parameters beta and mu in quadratic logistic models

Type "Dsoptimal" (without the double quotes) to run the program.

The model is .

- J. Qiu, R.-B. Chen, W. Wang, W. K. Wong (2012) .“Using Animal Instincts to Design Efficient Studies,” Technique report. [download]
- Forniusa, E. F. and Nyquista, H. (2009) Using the Canonical Design Space to Obtain c-optimal Designs for the Quadratic Logistic Model. Communications in Statistics - Theory and Methods, Volume 39, 144-157

**Part D. Locally designs for compartmental model**

The model is .

- D1: Locally D-optimal design (Compartmental model D-opt)

Type "run" (without the double quotes) to run the program. - D2: Locally c-optimal design for the area under curve (Compartmental model AUC)

Type "run" (without the double quotes) to run the program. - D3: The locally c-optimal design for the time to maximum concentration (Compartmental model Time to max con)

Type "run" (without the double quotes) to run the program.

- J. Qiu, R.-B. Chen, W. Wang, W. K. Wong (2012) .“Using Animal Instincts to Design Efficient Studies,” Technique report. [download]
- Atkinson, A. C., Donev, A. N. and Tobias, R. D. (2007). Optimum Experimental Designs, with SAS. Oxford Statistical Science series, Vol. 34. 261-264

**Part E. Locally D-optimal designs for 4-parameter Hill model**

- E1: Locally D-optimal designs for 4-parameter Hill model (Hill model D-opt)

Type "run" (without the double quotes) to run the program.

The model is .

- J. Qiu, R.-B. Chen, W. Wang, W. K. Wong (2012) .“Using Animal Instincts to Design Efficient Studies,” Technique report. [download]
- Leonid A., Khinkis, L. A., Levasseur, L. Faessel, H. and Greco, W. R. (2003). Optimal Design for Estimating Parameters of the 4-parameter Hill Model. Nonlinearity in Biology, Toxicology and Medicine, Vol.1 #3, 363-377.

**Part F. Locally D-optimal design for the double-exponential regrowth model**

- F1: Locally D-optimal design for the double-exponential regrowth model (Double Exp model D-opt)

Type "run" (without the double quotes) to run the program.

The model is .

- Li G. and Balakrishnan N. (2011) Optimal designs for tumor regrowth models. J. Statist. Plann. Inf., 141, 644-654.

**Part G. Two-parameter exponential model with type I right censored data**

The probability density function and the corresponding survival function are , where t is the observed value and x is the experimental condition for the observed data.

- G1: Locally D-optimal design (Exponential type I censoring D-opt for survival paper)

Type "run" (without the double quotes) to run the program. - G2: Locally c-optimal design for the parameter β (Exponential type I censoring c-opt for survival paper)

Type "run" (without the double quotes) to run the program.

- Konstantinou, M., Biedermann, S. and Kimber, A. (2011). Optimal designs for two-parameter nonlinear models with application to survival models. Southampton Statistical Research Institute, University of Southampton.

**Part H. Multivariate exponential and Poisson regression models**

For the kth dose combination, xk= (xk1, …, xkM)’, the mean of this M-variable model is assumed to be .

- H1: Locally D-optimal design for Poisson regression model with M = 2

Type "run" (without the double quotes) to run the program. - H2: Locally D-optimal design for Poisson regression model with M = 3

Type "run" (without the double quotes) to run the program. - H3: Locally D-optimal design for Poisson regression model with M = 4

Type "run" (without the double quotes) to run the program. - H4: Locally D-optimal design for Poisson regression model with M = 5

Type "run" (without the double quotes) to run the program. - H5: Locally D-optimal design for exponential regression model with M = 3

Type "run" (without the double quotes) to run the program. - H6: Locally D-optimal design for exponential regression model with M = 4

Type "run" (without the double quotes) to run the program. - H7: Locally D-optimal design for exponential regression model with M = 5

Type "run" (without the double quotes) to run the program.

**Disclaimer of Liability**

These programs are provided here for evaluating the capability of particle swarm optimization for finding optimal designs. The codes should be used as they are and without any warranty.