tradeoff.des {choiceDes}

MaxDiff and other tradeoff designs

Description

Generate a design to be used for MaxDiff and related tradeoff exercises.

Usage

tradeoff.des(items, shown, vers, tasks, fname=NULL, Rd=20, Rc=NULL, print=TRUE)

Arguments

items

The total number of items in the tradeoff exercise.

shown

The number of items shown in each tradeoff task.

vers

The number of blocks or versions in the final design.

tasks

The number of tradeoff tasks in each version of the final design.

fname

A character string, usually ending in ".txt", indicating the name of the file containing the tradeoff design.

Rd

The number of iterations in the design and blocking processes.

Rc

The number of iterations in the item by column position optimization routine.

print

Boolean indicating whether there is output to the console during execution.

Details

Replicates the functionality of Sawtooth Software MaxDiff Designer for designing MaxDiff and related tradeoff tasks.

A modified Federov (1972) algorithm is applied to a factor equal in length to the number of items to optimize the BIB design at vers x tasks rows and shown columns.

The optimized design is evaluated for one-way frequencies (equal representation of each item across all versions and column positions). Designs are also optimized for two-way or pairwise balance across all tasks. Column position balance is the more computationally-intensive process. The number of iterations required for this step is determined by the Rc argument which defaults to the larger of 1,000 or 10 x the number of rows in the design. Large design problems may require a larger number of iterations to achieve optimal column position balance.

Once an optimal design has been found, it is blocked into versions using optBlock to ensure equal representation of items within each version. See Wheeler (2004) for a more complete description of the modified Federov and blocking algorithms used in optimizing these designs.

If fname is not NULL a tab-delimited plain-text file is generated in the working directory containing the levels-coded design.

Value

design

A matrix consisting of the optimized design and additional variables for card, version and task.

balance

Tables of one-way item frequencies, two-way (pairwise) item frequencies, and item frequencies by column position. Means and standard deviations are calculated from all elements of the one-way and column position tables, and from the off-diagonal elements of the two-way (pairwise) table.

Rc.crit

The criterion that is minimized to achieve column position balance is output as a vector (crit.vec) along with the number of iterations executed since the last change in this criterion (crit.stable). If crit.stable is large or is a large proportion of the total number of iterations (Rc), the solution is stable in terms of column position balance. If crit.stable is small, the solution is likely unstable and column position balance could be improved by increasing Rc. See also cp.scree which produces a line plot of crit.vec as a function of Rc.

time.elapsed

The time required for the function to execute.

References

Federov, V.V. (1972). Theory of optimal experiments. Academic Press, New York.

Wheeler, R.E. (2004). AlgDesign. The R project for statistical computing. (http://www.r-project.org).

Examples

## Example 1:
## typical MaxDiff design with 12 items
des <- tradeoff.des(12, 4, 10, 9)

## Example 2:
## typical paired comparisons design with 14 items
des <- tradeoff.des(14, 2, 6, 14)