This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Result Response. Quantity of interest. Quality characteristic – Ntype, Stype etc.
2/1/2012
3
An example
Manufacturing process – air bubbles are identified. So rejection. Hoping to determine experimentally some way to fix the problem engineers identified 2 types of resins and 3 amts of prepolymers (100, 200, 300 gms) How would performance be measured
QC of the result
What are the first 2 factors to be identified and what are their levels
Result can be measured by counting number of bubbles (no units) or i f h b bbl ( )
Nlin applies only when continuous factors are considered
2/1/2012
5
Several factors at a time
One factor at a time – minimum two experiments
When multiple factors are present it is possible to run When multiple factors are present, it is possible to run N+1 experiments and get the same information of 2×N experiments effect
Though all factors are considered, its still one factor at a time. The way you conduct the experiment is different
2/1/2012
6
Experiments with multiple factors
Unfortunately, most real life problems have > 1 factor that influence the result and there is ‘interaction’that influence the result and there is interaction among the factors.
Once some responses are recorded based on a DOE -how do we go about making conclusions about thehow do we go about making conclusions about the system or the behavior?
We are interested in knowing which level or factor has the major influence on the
Consider a 1 factor multiple level problem with multiple experiments
Anova
We are interested in the mean of the observations within each level of our factor. The residuals will tell us about the variation within each level.
We can also average the means of each level to obtain a grand mean. We can then look at the deviation of the mean of each level from the grand mean to understand something about the level effects.
Finally, we can compare the variation within levels to the variation between levels.
Variability between the groups and variability within the groups.
Variability between the groups is calculated by first obtaining theVariability between the groups is calculated by first obtaining the sums of squares between groups (SSb), or the sum of the squared differences between each individual group mean from the grand mean.
Variability within the group is calculated by first obtaining the sums of squares within groups (SSw) or the sum of the squared differences between each individual score and that individual’s group mean
The diagrams below show the impact of increasing the numerator of the test statistic. Note that the within group variability (the denominator of th ti ) i th i it ti A d B H th b tthe equation) is the same in situations A and B. However, the betweengroup variability is greater in A than it is in B. This means that the F ratio for A will be larger than for B, and thus is more likely to be significant.
The diagrams below show the impact of decreasing the denominator of the test statistic. Note that the between group variability (the difference b t ) i th i it ti C d D H thbetween group means) is the same in situations C and D. However, the within group variability is greater in D than it is in C. This means that the F ratio for C will be larger than for D, and thus is more likely to be significant.
g ( ) Effect of ABC is not captured but remaining interaction(s) are captured Coeff of A is equivalent to BC ( B to AC and C to AB) This mix up is called Alias/confounding. Cant differentiate A, BC etc..
2/1/2012
18
How to find alias relationship
Defining relation I = ABC
Premultiply by A
=> A=BC, similarly for other combinations
2k-1 design is 1/2 fraction factorial design and is called the principal fraction
What happens to the second half of the table in the previous slide..?