Repository files navigation

The packageoptedr is an optimal experimental design suite forcalculating optimal designs, D-augmenting designs and efficientlyrounding approximate design. Among its capabilities are:

• Calculating D-, Ds-, A- and I-optimal designs for non-linear models.
• D-augment designs controlling the loss of efficiency.
• Evaluate the efficiency of a given design against the optimum.
• Efficiently rounding approximate designs to exact designs.

You can install the released version of optedr fromCRAN with:

install.packages("optedr")

You can install the latest version of the package fromGitHub with:

devtools::install_github("kezrael/optedr")

The user available functions are:

• opt_des() calculates optimal designs.
• design_efficiency() evaluates the efficiency of a design against theoptimum.
• augment_design() augments designs, allowing the user to add pointscontrolling the D-efficiency.
• efficient_round() efficiently round approximate designs.
• shiny_optimal() demo of optimal designs calculation with a graphicalinterface, applied toAntoine’s Equation.
• shiny_augment() demo of augmenting design with a graphicalinterface, usable for a handful of models.

Theoptdes object generated byopt_des() has its own implementationofprint(),summary() andplot().

library(optedr)

#> ℹ Loading optedr

The calculation of an optimal design requires a to specify thecriterion, themodel, theparameters and their initial values andthedesign_space.

resArr.D<- opt_des(criterion="D-Optimality",model=y~a*exp(-b/x),parameters= c("a","b"),par_values= c(1,1500),design_space= c(212,422))#>#> ℹ Stop condition not reached, max iterations performed#> ⠙ Calculating optimal design 22 done (27/s) | 812msℹ The lower bound for efficiency is 99.9986396401789%#>resArr.D$optdes#>      Point    Weight#> 1 329.2966 0.5000068#> 2 422.0000 0.4999932resArr.D$sens

resArr.D$convergence

After calculating the D-optimal design, the user might want to addpoints to the design to fit their needs:

aug_arr<- augment_design("D-Optimality",resArr.D$optdes,0.3,y~a* exp(-b/x),parameters= c("a","b"),par_values= c(1,1500),design_space= c(212,422),F)aug_arr#> [1] NA

This new design can be rounded to the desired number of points:

(exact_design<- efficient_round(aug_arr,20))

#>      Point Weight#> 1 329.2966      7#> 2 422.0000      7#> 3 260.0000      3#> 4 380.0000      3

And its efficiency compared against the optimum:

aprox_design<-exact_designaprox_design$Weight<-aprox_design$Weight/ sum(aprox_design$Weight)design_efficiency(aprox_design,resArr.D)#> ℹ The efficiency of the design is 86.0744360399533%#> [1] 0.8607444