In: Statistics and Probability
Manufacturing integrated circuits is an enormously complicated
task, as
there are many process variables that can be manipulated (thickness
of this,
width of that, doping level of something else, etc). Generally
speaking, many
copies of a circuit are put onto a single wafer, which is made all
at once. We
have an experiment where we are varying two factors, each at two
levels.
We have 20 wafers and assign each of the four factor/level
combinations to
five wafers at random and make the wafer. We then choose three
circuits at
random on each wafer and measure the performance of the circuit.
Construct
a skeleton ANOVA table (that is, just the sources and degrees of
freedom).
Soln
Factor 2 |
||
Factor 1 |
Level 1 |
Level 2 |
Level 1 |
1 |
1 |
2 |
2 |
|
3 |
3 |
|
4 |
4 |
|
5 |
5 |
|
Level 2 |
1 |
1 |
2 |
2 |
|
3 |
3 |
|
4 |
4 |
|
5 |
5 |
Null and Alternative Hypothesis:
We will have three hypotheses:
Factor 1
H0: µLevel 1 = µ Level 2
H1: Not all Means are equal
Factor 2
H0: µLevel 1 = µ Level 2
H1: Not all Means are equal
H0: An interaction is absent
H1: Interaction is present
Alpha = 0.05
Degress of Freedom:
DfFactor 1(A) = a-1 = 2-1 = 1
DfFactor 2 (B) = b-1 = 2-1 = 1
df Factor 1 * Factor 2 (A*B) = (a-1) * (b-1) = 1*1 = 1
df error = N – ab = 20 – 2*2 = 16
df total= N – 1 = 20 – 1 = 19
Critical Values:
Factor 1 (dfFactor 1(A), df error): (1,16) = 4.49
Factor 2 (dfFactor 2 (B),df error): (1,16) = 4.49
Interaction (df Factor 1 * Factor 2 (A*B), df error) : (1,16) = 4.49
[Factor 1] If F is greater than 4.49, reject the null hypothesis
[Factor 2] If F is greater than 4.49, reject the null hypothesis
[Interaction] If F is greater than 4.49, reject the null hypothesis
ANOVA Table Format
SS |
Df |
MS |
F |
|
Factor 1 |
1 |
|||
Factor 2 |
1 |
|||
Interaction |
1 |
|||
Error |
16 |
|||
Total |
19 |