Previous | Table of Contents | Next |
One way to find out the importance of a factor in a model is to find the percentage of variation explained by the factor. Another simpler informal alternative is to find the average response corresponding to each level of the factor and find the difference between the maximum and the minimum of such averages. This difference is called the range. A factor with a large range is considered important. It is assumed, though, that the observations have been transformed into a scale for which arithmetic averaging makes sense.
TABLE 23.10 Factor Averages and Range for the Paging Study | ||||
---|---|---|---|---|
Factor | Level 1 | Level 2 | Level 3 | Range of of Averages |
Replacement algorithm | 2056 | 2986 | 3781 | 1725 |
Deck arrangement | 1584 | 2913 | 4326 | 2742 |
Problem program | 592 | 2047 | 6185 | 5593 |
Memory size | 305 | 2006 | 6512 | 6207 |
The concepts of experimental design were developed for agriculture, and therefore, most of the examples in books on experimental design are from this held and not from computer science. It is therefore not surprising that this topic has been ignored in most textbooks on computer science performance analysis.
There are numerous books on design and analysis of experiments. In particular, see Mason, Gunst, and Hess (1989); Box, Hunter, and Hunter (1978); Dunn and Clark (1974); Hicks (1973); and Montgomery (1984). One of the recent developments in experimental design is the so-called Taguchi method. For details of this method see Ross (1988).
The scheduler design case study is from McHugh and Tzelnic (1981). The data for the code size comparison study of Examples 20.1 to 20.6 was adopted from that presented by Patterson and Sequin (1982). The data for Case Study 21.4 is from Hansen et al. (1982). The data as well as analysis for Case Study 23.1 is from Tsao and Margolin (1971).
Previous | Table of Contents | Next |