334
McCarville & Montgomery
Table 1. The Central Composite Design Ezperimental Settings for the Independent Yariables GDI and GB2, Coded from >1.414 to +1.414
FACTOR |
-1.414 |
-1 |
0 |
+ 1 |
+ 1.414 |
GB1 |
003.79 |
010 |
025 |
040 |
040.21 |
GB2 |
003.79 |
010 |
025 |
040 |
040.21 |
Table 2. The Results of Each of the Ninę Experimental Runs with Dependent Yariables ppm(Defective) and Gage Losses
Run # |
GB1 |
GB2 |
GB1 |
GB2 |
Defective |
Losses |
1 |
-1 |
-1 |
010 |
010 |
12.8958 |
107899 |
2 |
1 |
-1 |
040 |
010 |
59.2724 |
82328.8 |
3 |
-1 |
1 |
010 |
040 |
243.089 |
85939.8 |
4 |
1 |
1 |
040 |
040 |
1170.92 |
43897 |
5 |
-1.414 |
0 |
©03.79 |
025 |
400.0872 |
104093 |
0 |
1.414 |
0 |
040.21 |
025 |
413.008 |
55290.4 |
7 |
0 |
-1.414 |
025 |
003.79 |
13.0908 |
102.132 |
8 |
0 |
1.414 |
025 |
040.21 |
875.007 |
58290.3 |
9 |
0 |
0 |
025 |
025 |
107.034 |
72895.4 |
Notice that only one center point is chosen at this time. Since there is no random error in this experiment, any replication would be an exact copy of the initial center point. This will result in a pure error term of zero which in tum will erroneously inflate the F values for the model. However, as shown later, there are disadvantages to not replicating the center point.
These responses were analyzed with the Design-Expert software package, and response surface models were constmcted. For morę information