# SAS Macro for Charting the Relative Costs of Reduced Factorial Designs Public

The SAS macro RelativeCosts1 is provided as a supplement to Collins, Dziak, & Li (2009). It automates the calculation and graphing of relative costs of implementing different designs for multiple-factor experimental designs for screening or testing the effects of several independent dichotomous factors. It is written for use with SAS 9.1 for Windows. Sample File 1 can be run to demonstrate its use in a situation in which per-condition overhead costs are expected to be greater than per-subject marginal costs, and Sample File 2 demonstrates its use in the opposite situation.

## README

# SAS Macro for Charting the Relative Costs of Reduced Factorial Designs

## Overview

The SAS macro RelativeCosts1 is provided as a supplement to Collins, Dziak, & Li (2009). It automates the calculation and graphing of relative costs of implementing different designs for multiple-factor experimental designs for screening or testing the effects of several independent dichotomous factors. It is written for use with SAS 9.1 for Windows. Sample File 1 can be run to demonstrate its use in a situation in which per-condition overhead costs are expected to be greater than per-subject marginal costs, and Sample File 2 demonstrates its use in the opposite situation.

To use the macro, call it using the syntax below, replacing the "???"'s with the appropriate values.

%INCLUDE("c:\RelativeCosts1.sas");

%RelativeCosts1( number_of_factors=???, desired_fract_resolution=???, min_target_d_per_factor=???, condition_costlier_than_subject=???, max_cost_ratio=???, num_graph_points=???,max_graph_ratio=???);

The meaning of the arguments are listed below.

`Number_of_factors`

should be an integer between 2 and 19, telling the number k of factors of interest.`Desired_fract_resolution`

should be an integer between 3 and 6, giving the desired resolution of the fractional factorial design to be considered.`Min_target_d_per_factor`

is the effect size for which a power of .80 to detect is desired. This effect size refers to the main effect for a full or fractional factorial design, or to the appropriate simple effect in a single-factor-many-levels or many-separate-experiments design.`Condition_costlier_than_subject`

should be 0 if the per-condition overhead cost is expected to be less than the per-subject cost (e.g., if the cost of a treatment level is $500 plus $5000 per subject) and 1 if it is expected to be greater (e.g., if the cost of a treatment level is $5000 plus $500 per subject).`Max_cost_ratio`

,`num_graph_points`

, and`max_graph`

ratio control aspects of the appearance of the graph. These arguments can often be omitted, in which case they are left at sensible default levels.

*The macro prints information such as the following to the output window.*

- Doing separate experiments on each of the 9 factors requires at least: 7074 subjects total, i.e. 393 subjects in each of 18 cells.
- A comparative setup with 9 groups plus a control group requires at least: 3930 subjects total, i.e. 393 subjects in each of 10 cells.
- A 2^9 complete factorial experiment requires at least: 1024 subjects total, i.e. 2 subjects in each of 512 cells.
- A resolution 4 fractional factorial experiment with 9 factors requires at least: 800 subjects total, i.e. 25 subjects in each of 32 cells

The macro also produces a graph showing the relative costs of each design as a function of the ratio of per-condition overhead cost to per-subject cost (if condition*costlier*than*subject = 1) or per-subject cost to per-condition overhead cost (if condition*costlier*than*subject = 0). The ratios considered are 0 to max*cost*ratio, where the latter has a default value of 100. This macro is free and is provided for use without any warranty, implied or otherwise.

## Downloads

- Relative Costs SAS Macros:
`RelativeCosts1.zip`

- Example 1 for Using the Macro:
`RelativeCostsExample1.zip`

- Example 2 for Using the Macro):
`RelativeCostsExample2.zip`

## Recommended Citations

Collins, L. M., Dziak J. J., & Li, R. (2009). Design of experiments with multiple independent variables: A resource management perspective on complete and reduced factorial designs. *Psychological Methods, 14*(3), 202-24. PMCID: PMC2796056

## License

*See LICENSE.txt*

## Metadata

- Creator
- The Methodology Center, Penn State
- Keyword
- multiple-factor experiments
- SAS
- Rights
- All rights reserved
- Resource type
- Software or Program Code
- Published Date
- 2009
- Identifier
- https://doi.org/10.26207/3n6z-he59
- Size
- 9.2 KB
- Total items
- 5

## Collections

## Downloadable Content

Download file## Items in this Work

Title | Date Uploaded | Visibility | Actions | |
---|---|---|---|---|

LICENSE.txt | 2020-05-22 | Public | Download | |

README.txt | 2020-05-22 | Public | Download | |

RelativeCosts1.zip | 2020-05-22 | Public | Download | |

RelativeCostsExample1.zip | 2020-05-22 | Public | Download | |

RelativeCostsExample2.zip | 2020-05-22 | Public | Download |