A SAS PROGRAM FOR A 24 FULL FACTORIAL EXPERIMENT AT TWO LEVELS

1
A SAS PROGRAM FOR A 24 FULL FACTORIAL EXPERIMENT
AT TWO LEVELS

• EPITAXIAL LAYER GROWTH
• EXPERIMENT
• AMS 582
• Mickey Phagoo

2
Reference Text

• Experiments
• Planning, Analysis, and
• Parameter Design Optimization
• By C.F. Jeff Wu , Michael Hamada

3
Discussion

• full factorial design
• Specifically, how we would enter the code in SAS
• And the interpretation of the output.

4
EPITAXIAL LAYER GROWTH EXPERIMENT (PG 97 )
5
The experiment
6
How the data is entered in SAS

• The values in the first run and the corresponding
  16 runs each contain the four factor levels
  (A,B,C,D) and six replicates.
• Therefore, for the first run with the
  factors(-A,-B,-C,D) the first replicated value is
  entered , then the second until the sixth.

7
SAS PROGRAM
8
Using an infile instead of entering datelines
9
From the Print Function
10

• The syntax for finding the
• main effects and
• Performing the analysis

11
(No Transcript)
12
Temp Dataset B
13
MEAN, VARIANCE NATURAL LOG OF THE VARIANCE
14
Orthogonal Matrix
15
Factorial Effects of the Design
16
Factorial Effects of the Design ( )
17
Factorial Effects of Natural log of Variance
Regression Model
18
Factorial Effects of Natural log of Variance (ln
s2)
19
Regression Model ln s2
20
Regression Model for the Mean
21
The Model

• Therefore from the regression model, D and CD
  are significant .
• Note the purpose of the experiment is to find the
  process conditions i.e. the combinations of
  factor levels A B C and D under which the
  average thickness is close to the target value of
  14.5 microns with variation as small as possible
  and within an acceptable range of (14-15
  )microns.

22
Conclusion

• Therefore in conclusion weve learned how to
  enter the data as datalines , how to call an
  infile
• We spent some time on the actual syntax
• and the SAS output
• Weve determine a possible model using the main
  effects and other significant variables.
• Note you can streamline this code for different
  factorial designs and include additional code for
  doing further analysis as necessary.