Chronology | Current Month | Current Thread | Current Date |
[Year List] [Month List (current year)] | [Date Index] [Thread Index] | [Thread Prev] [Thread Next] | [Date Prev] [Date Next] |
I have seen it used and misused.
all of us who were active in the field knew what
the important processing variables already were
APPROACH 3: Pick a small number (often two) different choices for each
variable. You can then run all possible trials (2^6 = 64) - known as a
"full factorial design" - in our case. Or better, run a carefully chosen
subset - a "partial factorial" - as illustrated below. The subset of 16
listed below is not unique, but it does have special properties. Each
factor is high 8 times and low 8 times. Furthermore, in the set of 8
trials where any one variable is high, each other variables is high 4 times
and low 4 times. You lose some info by cutting the trials, but you save
time. You discover which variable have large effects and which have small
effects. You also get a good idea about which of the variables are
interacting with others. And note the random order, which cuts down on
unidentified systematic changes.
Run A B C D E F
1 - + - - + +
2 + + - - - +
3 + - - - + -
4 - + + - + -
5 + + + - - -
6 - - - + - -
7 - + - + + +
8 - + + + + -
9 + + - + - +
10 + - + + + +
11 + - - + + -
12 - - - - - -
13 + + + + - -
14 - - + - - +
15 - - + + - +
16 + - + - + +