View
1.646
Download
13
Tags:
Embed Size (px)
DESCRIPTION
Design of Experiments Lecture 2
Citation preview
1Design of Experiments
Simplified (DOES)
Presented By:Juanito S. Chan, PIE
2
Rationale
In any manufacturing setting, processes are governed by different variables that can affect the overall performance of the different processes. It is therefore necessary to identify these critical process variables with the purpose of controlling these variables in order to ensure a smooth process and to achieve the optimal operations of manufacturing.
3
Objectives
Identify and use the appropriate hypothesis test for comparing two sets of data
Identify and use the appropriate design for comparing more than two data sets
Design efficient experiments Quantify the effects of factors in the experiment Identify presence of interactions among the different
process variables Predict response as a function of the levels of the factors
4
Course Outline
Introduction Learning Process Role of Experimental Design Guidelines for Designing Experiments How to Use Statistical Techniques
Simple Comparative Experiments Inferences about the Difference in Means Inferences About the Variance Inferences About Proportions Contingency Tables and Test of Associations
5
Course Outline (Cont…)
Multiple Factor Experiments Introduction to Factorial Designs The 2k Factorial Design The General 2k-p Fractional Factorial Designs
6
Improvement Horizons
Customers
Design Optimization
Elimination
Process Optimization
Rectification
Process Control
Prevention
Product ControlDetection
7
DOE Potential Accomplishments
DOE
Enhance Common Cause Variability Management
Enhance Robustness of Processes and Products
Identify Complex Special Causes
Maintain Processes on Target
Reduce Variability
Increase Yield
Less Sensitive to Environmental Variation
Less Sensitive to InternalComponent Variation
Less Sensitive to “piece-to-piece” Variation
Applicable when SpecialCauses Involve Interactions
of Several Variables
8
SPC/DOE Roles in Improvement Efforts
SPCPassive Observation
DOEParticipative Observation
Informed observation ofnaturally OccurringInformative Events
Perturbing Process toInvite Occurrence ofInformative Events
Increased Probabilityof Observing KeyEvents when theyNaturally Occur
Increased Probability ofKey Events Occurring
9
Learning Process
Data (facts, phenomena)
Hypothesis (conjecture, model, theory)
deduction
induction induction induction
deduction deduction
... ...
10
How to Use Statistical Techniques
Use your non-statistical knowledge of the problem
Keep the design and analysis as simple as possible
Recognize the difference between practical and statistical significance
Experiments are usually iterative
11
Three R’s of DOE
andomization sequence of experiments and/or the
assignment of specimens to various treatment combinations in a purely chance manner
eplication infers two or more runs were conducted under
the same test conditions, each run following a new set-up or resetting of the conditions
epetition obtaining more than one measurement or unit
of output for each run
12
Obtain a clear statement of the problem Identify the problem area in quantitative terms Identify the response(s) to be measured, the factors that may
be varied, the factors to be held constant, and the factors that cannot be controlled
Identify the ranges or limitations of the measurements and of the experimental factors
Collect available background information Investigate all available sources of information Tabulate data pertinent to planning the experimental program Be quantitative
Checklist for Planning Experiments
13
Checklist for Planning Experiments Design the experimental program
Hold a conference of all parties concerned State the propositions to be explored Agree on magnitude of differences in the response considered
worthwhile Outline possible alternative outcomes Choose the factors to be studied Determine practical range of factors and specify levels Choose the measurements and methods of measurement Consider the effect of sampling variability and of precision of the
measurement methods Consider possible interrelationships of the factors Determine influences of time, cost, materials, manpower,
instrumentation, and other facilities and of extraneous conditions such as weather
Consider personnel and human relations requirements of the program
14
Design the experimental program in preliminary form Prepare a systematic and inclusive schedule, which includes
the randomization pattern Provide for stepwise performance or adaptation of schedule if
necessary Eliminate effect of variables not under study by controlling,
balancing, or randomization Minimize the number of experimental runs consistent with
objectives Choose the method of statistical analysis Arrange for orderly accumulation of data Review the experimental design program with all concerned Adjust the program as required Spell out the steps to be followed in unmistakable terms
Checklist for Planning Experiments
15
Plan and carry out the experimental work Develop methods, materials, and equipment Carry out the experimental design in some random order Record ancillary data Record any modifications of the experimental design Take precautions in the collection and recording of data,
especially data from extra experiments and missing experiments Record progress of the program by date, run number, and other
ancillary data Analyze the data
Review the data with attention to recording errors, omissions, etc.
Use graphics: plot the data, plot averages, plot simple graphs Apply appropriate statistical techniques
Checklist for Planning Experiments
16
Interpret the results Consider all the observed data Confine initial conclusions to strict deductions from the
experimental evidence at hand Elucidate the analysis in both graphical and numerical terms State results in terms of verifiable probabilities Arrive at conclusions as to the technical meaning of results
as well as their statistical significance Point out implications of the findings for application and for
further work Account for any limitations imposed by the data or by the
methods of analysis used
Checklist for Planning Experiments
17
Prepare the report Describe work clearly, giving background, pertinence of
problems, meaning of results Use tabular and graphic methods of presenting data, and
consider their possible future use Supply sufficient information to permit readers to verify
results and to draw their own conclusions Limit conclusion to objective summary of evidence
Checklist for Planning Experiments
18
DOE Vocabulary Factor
One of the independent variables under investigation that can be set to a desired value
k the number of factors or
variables, the effects of which are to be estimated in an experiment
Level the numerical value or
qualitative feature of a factor
Run the act of operating the
process with the factors at certain settings
Treatment specific combination of
the levels of all factors for a given test or run
Response the numerical result of a
run
19
DOE Vocabulary Experimental error
the amount of variability that may be expected in the experimental environment just by chance without any changing of the factors being investigated
Main Effect the average influence on the
response as a variable changes levels
Interaction Effect the average difference in the
effect on a response of one variable dependent upon the settings of another variable
MSFE Minimum Significant Factor
Effect is the minimum absolute value of an effect which may be considered a significant result
Factorial experiment designed to determine the effect
of all possible combinations across all levels of the factors under study
Fractional Factorial designed to examine k factors
with a fraction of the runs required for a full factorial
20
DOE Vocabulary Confounding
the consequences of conducting a fractional factorial design
Blocking a strategy for designing experiments
to provide the ability to eliminate from the experimental error a contributor of variability that is known but not under investigation
Robust the quality of a process or output
being little affected by environmental or internal component variation
Noise refers to variability, frequently
uncontrollable or random variability in experimental design work
ANOVA a mathematical procedure
testing for significant differences between or among groups
21
Strategies for Designing Experiments
Screening
Highly fractionalizedfactorials
Many potentially important
variables Few tests
Knowledge BuildingStage
Full Factorials or Highresolution fractional
factorials
A few important variablesemerge
Optimization
Replicated designsFull factorials
Evolutionary OperationResponse Surface
Methodology
Best settings are determined
22
Simple Comparative Experiments
23
Types of Errors
Type I Error ( risk) Reject the hypothesis when it
is true Type II Error ( risk)
Accept the hypothesis when it is false
24
Steps in Hypothesis Testing State the hypothesis Choose the Type I error Choose the test statistic for testing the
hypothesis Determine the acceptance region for the test Obtain the sample of observations, compute the
test statistic, and compare the value to the acceptance region to make a decision to accept or reject the hypothesis
Draw an engineering conclusion
25
Summary Table of Hypotheses Tests
n
xZ
/0
X is distribution-free but shouldbe continuous and have only onemode
Standard deviation ofpopulation is known
Normal distribution
ns
xt
/0
X is normally distributed Standard deviation of
population is estimatedby sample s
Test 1. The mean of a population is equal to 0 (Ho: = 0)
26
Summary Table of Hypotheses Tests
2
22
1
21
21
nn
xxZ
X1 and X2 are distribution-free but shouldbe continuous and have only one mode. Ifthe populations are not normally distributed,sample sizes n1 and n2 should be large so that sampling distribution of Z is approximatelynormal
Standard deviations of populations are known
Normal distribution
21
21
11nn
s
xxt
p
1 = 2. X1 and X2 are normally distributed Standard deviations ofpopulations are estimatedby sample s1 and s2
Test 2. The means of two populations are equal (Ho: 1 = 2)
2
11
21
222
2112
nn
snsnsp
t distribution with df = n1 + n2 - 2
27
Summary Table of Hypotheses Tests
ns
dt
d /
Populations are normally distributed Data are taken in n pairsand difference d withineach pair is calculated
t distribution with df = n - 1
2
22
1
21
21
ns
ns
xxt
X1 and X2 are normally distributed Standard deviations ofpopulations are estimatedby sample s1 and s2 (noassumption that 1 = 2)
Test 2. The means of two populations are equal (Ho: 1 = 2)
t distribution with df = min(n1 - 1, n2 - 1)
28
Summary Table of Hypotheses Tests
20
22 1
sn
Populations are normally distributed Standard deviation ofpopulation is estimatedby sample s
Chi-square dist. With df = n - 1
Test 3. The standard deviation of a population is equal to 0 (Ho: = 0)
29
Summary Table of Hypotheses Tests
22
21
s
sf
Populations are normally distributed Standard deviation ofpopulation is estimatedby sample s1 and s2
F distribution with df1 = n1 - 1df2 = n2 - 1
Test 4. The standard deviations of two populations are equal (Ho: 12 = 2
2)
30
Summary Table of Hypotheses Tests
00
0
1 pnp
npXz
n > 100. Only for large sample sizes Proportion of population
is estimated by sampleproportion
Normal distribution
Test 5. The proportion of a population exhibiting a certain characteristic is po (Ho: p = po)
31
Summary Table of Hypotheses Tests
21
2211
/1/1ˆ1ˆ
//
nnpp
nXnXz
np > 5 for each population.Sample sizes n1 and n2 must be large so that sampling dist.of Z is approximately normal
Proportions in populationsare estimated by sampleproportions
Normal distribution
Test 6. The proportion in two populations are equal (Ho: p1 = p2)
21
21ˆnn
XXp
where
32
Example #1:
The shelf life of a carbonated beverage is of interest. Ten bottles are randomly selected and tested, and the following results are obtained:
Days: 108, 124, 124, 106, 115, 138, 163, 159, 134, 139
Assume that the alternative hypothesis is that the mean shelf life is greater than 125 days. Can the null hypothesis H0: = 125 be rejected?
33
Exercise #1:
The time to repair an electronic instrument is a normally distributed random variable measured in hours. The repair times for 16 such instruments chosen at random are as follows:
Hours
159, 280, 101, 212, 224, 379, 179, 264
222, 362, 168, 250, 149, 260, 485, 170
Does it seem reasonable that the true mean repair time is greater than 225 hours?
34
Example #2
A chemical engineer is investigating the inherent variability of two types of test equipment that can be used to monitor the output of a production process. He suspects that the old equipment, type 1, has a larger variance than the new one. Thus, he wishes to test the hypothesis
Ho:12 = 2
2
Ha:12 > 2
2
Two random samples of n1 = 12 and n2 = 10 observations are taken, and the sample variances are s1
2 = 14.5 and s22
= 10.8
35
Exercise #2
A machined engine part produced by a company is claimed to have diameter variance no larger than .0002”. A random sample of 10 parts gave a sample variance of .0003”. Test at the 5 percent level, Ho: 2 = .0002 against Ha: 2 > .0002.
36
Example #3 Two machines are used for filling plastic bottles with a net volume of
16.0 ounces. The filling processes can be assumed to be normal, with standard deviations of s1 = 0.015 and s2 = 0.018. The quality engineering department suspects that both machines fill to the same net volume, whether or not this volume is 16.0 ounces. A random sample is taken from the output of each machine.
Machine 1: 16.03 16.04 16.05 16.05 16.02 16.01 15.66 15.98 16.02 15.99
Machine 2: 16.02 15.97 15.96 16.01 15.99 16.03 16.04 16.02 16.01 16.00
Do you think the quality engineering department is correct?
37
Exercise #3
An article in Solid State Technology, “Orthogonal Design for Process Optimization and Its Application to Plasma Etching” by GZ Yin and DW Jillie (May, 1987) describes an experiment to determine the effect of the C2F6 flow rate on the uniformity of the etch on a silicon wafer used in integrated circuit manufacturing. Data for two flow rates are as follows:
C2F6 Flow Uniformity Observation
125 2.7 4.6 2.6 3.0 3.2 3.8
200 4.6 3.4 2.9 3.5 4.1 5.1
Does the C2F6 flow rate affect the wafer-to-wafer variability and average etch uniformity?
38
Example #4
An article in the Journal of Strain Analysis (vol. 18, no. 2, 1983) compares several procedures for predicting the shear strength for steep plate girders. Data for nine girders in the form of the ratio of predicted to observed load for two of these procedures, the Karlsruhe and Lehigh methods, are as follows:
Girder 1 2 3 4 5 6 7 8 9
KM 1.186 1.151 1.322 1.339 1.200 1.402 1.365 1.537 1.559
LM 1.061 0.992 1.063 1.062 1.065 1.178 1.037 1.086 1.052
Is there any difference between the two methods?
39
Exercise #4
The diameter of a ball bearing was measured by 12 inspectors, each using two different kinds of calipers. The results were
Insp 1 2 3 4 5 6 7 8
9 10 11 12
Caliper 1 0.265 0.265 0.266 0.267 0.267 0.265 0.267 0.267
0.265 0.268 0.268 0.265
Caliper 2 0.264 0.265 0.264 0.266 0.267 0.268 0.264 0.265
0.265 0.267 0.268 0.269
Is there a significant difference between the means of the population of measurements represented by the two samples? Use = .05
40
Statistical Tables
41
Table 1: t distribution0.1 0.05 0.025 0.01 0.005 0.0025
3.078 6.314 12.706 31.821 63.656 127.3211.886 2.920 4.303 6.965 9.925 14.0891.638 2.353 3.182 4.541 5.841 7.4531.533 2.132 2.776 3.747 4.604 5.5981.476 2.015 2.571 3.365 4.032 4.7731.440 1.943 2.447 3.143 3.707 4.3171.415 1.895 2.365 2.998 3.499 4.0291.397 1.860 2.306 2.896 3.355 3.8331.383 1.833 2.262 2.821 3.250 3.6901.372 1.812 2.228 2.764 3.169 3.5811.363 1.796 2.201 2.718 3.106 3.4971.356 1.782 2.179 2.681 3.055 3.4281.350 1.771 2.160 2.650 3.012 3.3721.345 1.761 2.145 2.624 2.977 3.3261.341 1.753 2.131 2.602 2.947 3.2861.337 1.746 2.120 2.583 2.921 3.2521.333 1.740 2.110 2.567 2.898 3.2221.330 1.734 2.101 2.552 2.878 3.1971.328 1.729 2.093 2.539 2.861 3.1741.325 1.725 2.086 2.528 2.845 3.1531.323 1.721 2.080 2.518 2.831 3.1351.321 1.717 2.074 2.508 2.819 3.1191.319 1.714 2.069 2.500 2.807 3.1041.318 1.711 2.064 2.492 2.797 3.0911.316 1.708 2.060 2.485 2.787 3.0781.315 1.706 2.056 2.479 2.779 3.0671.314 1.703 2.052 2.473 2.771 3.0571.313 1.701 2.048 2.467 2.763 3.0471.311 1.699 2.045 2.462 2.756 3.0381.310 1.697 2.042 2.457 2.750 3.030
123456789
101112131415161718192021222324
2930
25262728
v
42
Table 2: z distribution
z 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.090.0 0.50000 0.50399 0.50798 0.51197 0.51595 0.51994 0.52392 0.52790 0.53188 0.535860.1 0.53983 0.54380 0.54776 0.55172 0.55567 0.55962 0.56356 0.56749 0.57142 0.575350.2 0.57926 0.58317 0.58706 0.59095 0.59483 0.59871 0.60257 0.60642 0.61026 0.614090.3 0.61791 0.62172 0.62552 0.62930 0.63307 0.63683 0.64058 0.64431 0.64803 0.651730.4 0.65542 0.65910 0.66276 0.66640 0.67003 0.67364 0.67724 0.68082 0.68439 0.687930.5 0.69146 0.69497 0.69847 0.70194 0.70540 0.70884 0.71226 0.71566 0.71904 0.722400.6 0.72575 0.72907 0.73237 0.73565 0.73891 0.74215 0.74537 0.74857 0.75175 0.754900.7 0.75804 0.76115 0.76424 0.76730 0.77035 0.77337 0.77637 0.77935 0.78230 0.785240.8 0.78814 0.79103 0.79389 0.79673 0.79955 0.80234 0.80511 0.80785 0.81057 0.813270.9 0.81594 0.81859 0.82121 0.82381 0.82639 0.82894 0.83147 0.83398 0.83646 0.838911.0 0.84134 0.84375 0.84614 0.84849 0.85083 0.85314 0.85543 0.85769 0.85993 0.862141.1 0.86433 0.86650 0.86864 0.87076 0.87286 0.87493 0.87698 0.87900 0.88100 0.882981.2 0.88493 0.88686 0.88877 0.89065 0.89251 0.89435 0.89617 0.89796 0.89973 0.901471.3 0.90320 0.90490 0.90658 0.90824 0.90988 0.91149 0.91308 0.91466 0.91621 0.917741.4 0.91924 0.92073 0.92220 0.92364 0.92507 0.92647 0.92785 0.92922 0.93056 0.931891.5 0.93319 0.93448 0.93574 0.93699 0.93822 0.93943 0.94062 0.94179 0.94295 0.944081.6 0.94520 0.94630 0.94738 0.94845 0.94950 0.95053 0.95154 0.95254 0.95352 0.954491.7 0.95543 0.95637 0.95728 0.95818 0.95907 0.95994 0.96080 0.96164 0.96246 0.963271.8 0.96407 0.96485 0.96562 0.96638 0.96712 0.96784 0.96856 0.96926 0.96995 0.970621.9 0.97128 0.97193 0.97257 0.97320 0.97381 0.97441 0.97500 0.97558 0.97615 0.976702.0 0.97725 0.97778 0.97831 0.97882 0.97932 0.97982 0.98030 0.98077 0.98124 0.98169
43
Table 2: z distribution (cont…)
z 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.092.1 0.98214 0.98257 0.98300 0.98341 0.98382 0.98422 0.98461 0.98500 0.98537 0.985742.2 0.98610 0.98645 0.98679 0.98713 0.98745 0.98778 0.98809 0.98840 0.98870 0.988992.3 0.98928 0.98956 0.98983 0.99010 0.99036 0.99061 0.99086 0.99111 0.99134 0.991582.4 0.99180 0.99202 0.99224 0.99245 0.99266 0.99286 0.99305 0.99324 0.99343 0.993612.5 0.99379 0.99396 0.99413 0.99430 0.99446 0.99461 0.99477 0.99492 0.99506 0.995202.6 0.99534 0.99547 0.99560 0.99573 0.99585 0.99598 0.99609 0.99621 0.99632 0.996432.7 0.99653 0.99664 0.99674 0.99683 0.99693 0.99702 0.99711 0.99720 0.99728 0.997362.8 0.99744 0.99752 0.99760 0.99767 0.99774 0.99781 0.99788 0.99795 0.99801 0.998072.9 0.99813 0.99819 0.99825 0.99831 0.99836 0.99841 0.99846 0.99851 0.99856 0.998613.0 0.99865 0.99869 0.99874 0.99878 0.99882 0.99886 0.99889 0.99893 0.99896 0.999003.1 0.99903 0.99906 0.99910 0.99913 0.99916 0.99918 0.99921 0.99924 0.99926 0.999293.2 0.99931 0.99934 0.99936 0.99938 0.99940 0.99942 0.99944 0.99946 0.99948 0.999503.3 0.99952 0.99953 0.99955 0.99957 0.99958 0.99960 0.99961 0.99962 0.99964 0.999653.4 0.99966 0.99968 0.99969 0.99970 0.99971 0.99972 0.99973 0.99974 0.99975 0.999763.5 0.99977 0.99978 0.99978 0.99979 0.99980 0.99981 0.99981 0.99982 0.99983 0.999833.6 0.99984 0.99985 0.99985 0.99986 0.99986 0.99987 0.99987 0.99988 0.99988 0.999893.7 0.99989 0.99990 0.99990 0.99990 0.99991 0.99991 0.99992 0.99992 0.99992 0.999923.8 0.99993 0.99993 0.99993 0.99994 0.99994 0.99994 0.99994 0.99995 0.99995 0.999953.9 0.99995 0.99995 0.99996 0.99996 0.99996 0.99996 0.99996 0.99996 0.99997 0.999974.0 0.99997 0.99997 0.99997 0.99997 0.99997 0.99997 0.99998 0.99998 0.99998 0.99998
44
Table 3: f distribution
1 2 3 4 5 6 7 8 9 10 12 15 20 24 30 40 60 1201 161.45 199.50 215.71 224.58 230.16 233.99 236.77 238.88 240.54 241.88 243.90 245.95 248.02 249.05 250.10 251.14 252.20 253.252 18.51 19.00 19.16 19.25 19.30 19.33 19.35 19.37 19.38 19.40 19.41 19.43 19.45 19.45 19.46 19.47 19.48 19.493 10.13 9.55 9.28 9.12 9.01 8.94 8.89 8.85 8.81 8.79 8.74 8.70 8.66 8.64 8.62 8.59 8.57 8.554 7.71 6.94 6.59 6.39 6.26 6.16 6.09 6.04 6.00 5.96 5.91 5.86 5.80 5.77 5.75 5.72 5.69 5.665 6.61 5.79 5.41 5.19 5.05 4.95 4.88 4.82 4.77 4.74 4.68 4.62 4.56 4.53 4.50 4.46 4.43 4.406 5.99 5.14 4.76 4.53 4.39 4.28 4.21 4.15 4.10 4.06 4.00 3.94 3.87 3.84 3.81 3.77 3.74 3.707 5.59 4.74 4.35 4.12 3.97 3.87 3.79 3.73 3.68 3.64 3.57 3.51 3.44 3.41 3.38 3.34 3.30 3.278 5.32 4.46 4.07 3.84 3.69 3.58 3.50 3.44 3.39 3.35 3.28 3.22 3.15 3.12 3.08 3.04 3.01 2.979 5.12 4.26 3.86 3.63 3.48 3.37 3.29 3.23 3.18 3.14 3.07 3.01 2.94 2.90 2.86 2.83 2.79 2.7510 4.96 4.10 3.71 3.48 3.33 3.22 3.14 3.07 3.02 2.98 2.91 2.85 2.77 2.74 2.70 2.66 2.62 2.5811 4.84 3.98 3.59 3.36 3.20 3.09 3.01 2.95 2.90 2.85 2.79 2.72 2.65 2.61 2.57 2.53 2.49 2.4512 4.75 3.89 3.49 3.26 3.11 3.00 2.91 2.85 2.80 2.75 2.69 2.62 2.54 2.51 2.47 2.43 2.38 2.3413 4.67 3.81 3.41 3.18 3.03 2.92 2.83 2.77 2.71 2.67 2.60 2.53 2.46 2.42 2.38 2.34 2.30 2.2514 4.60 3.74 3.34 3.11 2.96 2.85 2.76 2.70 2.65 2.60 2.53 2.46 2.39 2.35 2.31 2.27 2.22 2.1815 4.54 3.68 3.29 3.06 2.90 2.79 2.71 2.64 2.59 2.54 2.48 2.40 2.33 2.29 2.25 2.20 2.16 2.1116 4.49 3.63 3.24 3.01 2.85 2.74 2.66 2.59 2.54 2.49 2.42 2.35 2.28 2.24 2.19 2.15 2.11 2.0617 4.45 3.59 3.20 2.96 2.81 2.70 2.61 2.55 2.49 2.45 2.38 2.31 2.23 2.19 2.15 2.10 2.06 2.0118 4.41 3.55 3.16 2.93 2.77 2.66 2.58 2.51 2.46 2.41 2.34 2.27 2.19 2.15 2.11 2.06 2.02 1.9719 4.38 3.52 3.13 2.90 2.74 2.63 2.54 2.48 2.42 2.38 2.31 2.23 2.16 2.11 2.07 2.03 1.98 1.9320 4.35 3.49 3.10 2.87 2.71 2.60 2.51 2.45 2.39 2.35 2.28 2.20 2.12 2.08 2.04 1.99 1.95 1.90
v1v2
45
Table 3: f distribution
1 2 3 4 5 6 7 8 9 10 12 15 20 24 30 40 60 12021 4.32 3.47 3.07 2.84 2.68 2.57 2.49 2.42 2.37 2.32 2.25 2.18 2.10 2.05 2.01 1.96 1.92 1.8722 4.30 3.44 3.05 2.82 2.66 2.55 2.46 2.40 2.34 2.30 2.23 2.15 2.07 2.03 1.98 1.94 1.89 1.8423 4.28 3.42 3.03 2.80 2.64 2.53 2.44 2.37 2.32 2.27 2.20 2.13 2.05 2.01 1.96 1.91 1.86 1.8124 4.26 3.40 3.01 2.78 2.62 2.51 2.42 2.36 2.30 2.25 2.18 2.11 2.03 1.98 1.94 1.89 1.84 1.7925 4.24 3.39 2.99 2.76 2.60 2.49 2.40 2.34 2.28 2.24 2.16 2.09 2.01 1.96 1.92 1.87 1.82 1.7726 4.23 3.37 2.98 2.74 2.59 2.47 2.39 2.32 2.27 2.22 2.15 2.07 1.99 1.95 1.90 1.85 1.80 1.7527 4.21 3.35 2.96 2.73 2.57 2.46 2.37 2.31 2.25 2.20 2.13 2.06 1.97 1.93 1.88 1.84 1.79 1.7328 4.20 3.34 2.95 2.71 2.56 2.45 2.36 2.29 2.24 2.19 2.12 2.04 1.96 1.91 1.87 1.82 1.77 1.7129 4.18 3.33 2.93 2.70 2.55 2.43 2.35 2.28 2.22 2.18 2.10 2.03 1.94 1.90 1.85 1.81 1.75 1.7030 4.17 3.32 2.92 2.69 2.53 2.42 2.33 2.27 2.21 2.16 2.09 2.01 1.93 1.89 1.84 1.79 1.74 1.6840 4.08 3.23 2.84 2.61 2.45 2.34 2.25 2.18 2.12 2.08 2.00 1.92 1.84 1.79 1.74 1.69 1.64 1.5860 4.00 3.15 2.76 2.53 2.37 2.25 2.17 2.10 2.04 1.99 1.92 1.84 1.75 1.70 1.65 1.59 1.53 1.47
120 3.92 3.07 2.68 2.45 2.29 2.18 2.09 2.02 1.96 1.91 1.83 1.75 1.66 1.61 1.55 1.50 1.43 1.35
v1v2
46
Table 4: Chi-square Dist.1.99 1.98 1.95 1.9 0.1 0.01
0.995 0.99 0.975 0.95 0.05 0.025 0.01 0.005
0.000 0.000 0.001 0.004 3.841 5.024 6.635 7.8790.010 0.020 0.051 0.103 5.991 7.378 9.210 10.5970.072 0.115 0.216 0.352 7.815 9.348 11.345 12.8380.207 0.297 0.484 0.711 9.488 11.143 13.277 14.8600.412 0.554 0.831 1.145 11.070 12.832 15.086 16.7500.676 0.872 1.237 1.635 12.592 14.449 16.812 18.5480.989 1.239 1.690 2.167 14.067 16.013 18.475 20.2781.344 1.647 2.180 2.733 15.507 17.535 20.090 21.9551.735 2.088 2.700 3.325 16.919 19.023 21.666 23.5892.156 2.558 3.247 3.940 18.307 20.483 23.209 25.1882.603 3.053 3.816 4.575 19.675 21.920 24.725 26.7573.074 3.571 4.404 5.226 21.026 23.337 26.217 28.3003.565 4.107 5.009 5.892 22.362 24.736 27.688 29.8194.075 4.660 5.629 6.571 23.685 26.119 29.141 31.3194.601 5.229 6.262 7.261 24.996 27.488 30.578 32.8015.142 5.812 6.908 7.962 26.296 28.845 32.000 34.2675.697 6.408 7.564 8.672 27.587 30.191 33.409 35.7186.265 7.015 8.231 9.390 28.869 31.526 34.805 37.1566.844 7.633 8.907 10.117 30.144 32.852 36.191 38.5827.434 8.260 9.591 10.851 31.410 34.170 37.566 39.9978.034 8.897 10.283 11.591 32.671 35.479 38.932 41.4018.643 9.542 10.982 12.338 33.924 36.781 40.289 42.7969.260 10.196 11.689 13.091 35.172 38.076 41.638 44.1819.886 10.856 12.401 13.848 36.415 39.364 42.980 45.558
10.520 11.524 13.120 14.611 37.652 40.646 44.314 46.92811.160 12.198 13.844 15.379 38.885 41.923 45.642 48.29011.808 12.878 14.573 16.151 40.113 43.195 46.963 49.64512.461 13.565 15.308 16.928 41.337 44.461 48.278 50.99413.121 14.256 16.047 17.708 42.557 45.722 49.588 52.33513.787 14.953 16.791 18.493 43.773 46.979 50.892 53.672
123456789
101112131415161718192021222324
2930
25262728
v
47
B vs C Experiment
48
A quick and simple non-parametric method of comparing two samples and deciding whether one is better than the other
B is the code used for “Better” and C is the code for “Current”
What is B vs. C?
49
Choose an acceptable level of risk ()
Decide on Sample Sizes for B and C Tests
Randomize and Conduct the Tests
Rank Order the Results
Decision Rule No Overlap With Overlap (End-Count) Technique
General Procedure: B vs. C
50
Choose an acceptable level of risk ()
Decide on Sample Sizes for B and C Tests (Base on the Table for B vs. C With Overlap)
Randomize and Conduct the Tests
Rank Order the Results
Decision Rule SIMPLE RULE: An entity is better than the other if all
readings outrank all readings of the other entity.
B vs. C: No Overlap Rule
51
B vs. C: Table for No Overlap Rule
52
A process change on a tuning coil was expected to increase yield. It was decided to list yields of three C lots and four B lots. It was stated that the proposed change would only be put into effect if all B yields outranked all C yields
C: 93.6 93.8 92.5 B: 94.1 94.3 93.7 94.2
Exercise 1
53
Choose an acceptable level of risk () Decide on Sample Sizes for B and C Tests
Tukey’s Overlap End Count Technique: Sample sizes are larger, generally 10 or more for each If sample sizes are unequal, the ratio of nB:nC should be no
more than 3:4 Randomize and Conduct the Tests Rank Order the Results Decision Rule -risk 0.10 0.05 0.01 0.001
End Count 6 7 10 13> =
B vs. C: With Overlap Rule
54
What is an “END-COUNT” ?
BBBBBB CBBBCCCCBBCB CCCCCCCC
“high” end “low” endOVERLAP
END-COUNT = #high end + #low end = 6 + 8 = 14
B vs. C: With Overlap Rule
55
What is an “END-COUNT” ?
CCCCCC BBBBCCCCBBCC BBBBBBB
“high” end “low” endOVERLAP
END-COUNT = 6 + 7 = 13
CCCCCC BBBBCBBCBBCCBBBCCCC
“high” end OVERLAP
END-COUNT = 6 + 0 = 6
B vs. C: With Overlap Rule
56
In the fabrication of a 64K RAM, a B vs C test was run to see if standard substrates in a room atmosphere (C) could be produced in a high-oxygen atmosphere (B) in order to improve yields. Twelve C and thirteen B samples were selected and processed in random order C: 95.6 92.5 98.5 94.6 95.8 88.3 94.1 90.5 97.5 94.9
93.7 90.0 B: 96.7 91.2 99.2 98.6 97.0 99.4 93.6 97.2 94.5 93.2
99.3 90.4 98.2
Exercise 2
57
Factorial Experiments
58
One-level Factorial Designs
One-level factorial designs are experimental designs in which each of the k factors is set at a single level only for all the test combinations that are conducted.
They are similar to one-way ANOVA. They are very useful for simple comparisons of categorical treatments.
59
General one factor design problem
Three bowlers must compete for the last position in the national team. They bowl six games. (see data on the next slide)
The captain knows better to just simply pick the bowler with highest score. Maybe it’s a fluke that Mark scored highest and Pat’s score is low. He wants to know if the scores are significantly different, given the variability in individual scores.
60
Game Pat Mark Shari
1 160 165 166
2 150 180 158
3 140 170 145
4 167 185 161
5 157 195 151
6 148 175 156
61
The bowling captain does not care if averages differ by less than 10 pins and his records provide the standard deviation of 5.
Which of the three bowlers should he pick for the national team?
Justify your conclusion using the Design Expert software.
62
Two-level Factorial Designs
Two Level factorial designs are experimental designs in which each of the k factors is set at one of two levels (high or low) all 2k test combinations of factor levels are conducted
A 22 Full Factorial Design
A B1 - +2 + -3 - -4 + +
63
Generalized 2k Designs
64
Estimation of Main Effects
A main effect (or average main effect) of a factor is the difference between the response (or average of several responses) at the high level of the factor and the response (or average of several responses) at the low level of the factor
92
93
94
95
96
97
98
99
- +
Y
Main effect
65
Calculation Matrix
Run X1 X2 X1X2
1 - - +
2 + - -
3 - + -
4 + + +
Factor
Design matrix
Calculation matrix
List all possible interactions
Multiply the row entries of the respective columns associated with the variables that are listed in the interaction column heading
Calculation of effect estimates is similar to the main effects calculation
66
Graphical Aids for Analysis
92
93
94
95
96
97
98
99
- +
Y
90
92
94
96
98
100
102
- +
X2 (low) X2 (high)
One Factor Main Effect Two Factor Interaction Effect
67
General Nature of Interactions
No Interaction
No Interaction
SlightInteraction
SlightInteraction
StrongInteraction(reversal)
StrongInteraction(reversal)
The interaction plot is a wayto visualize the
inter-dependency betweenfactors
The Calculationsare not difficultthe MEANING
is critical
Low High
HighLow
B
Low High
HighLow
B
Low High
HighLow
BBB
A-A-
BB
BB
A+A+
A-A-
A+A+
A-A-
A+A+
0102030
40506070
0102030
40506070
0102030
40506070
68
Steps in Judging Relative Importance of Effects
Estimate experimental error (s) Determine the standard error of effects State the precision for an effect as
a confidence interval about the effect estimates a range of values centered at zero which would
have happened by chance
69
Estimating Experimental Error for Location Effects
i
ii
v
sv 2
N
ss pe
2
N
Es ie
2
estimate graphical es
Normal Probability Plotting(NPP)
Replicated Experiment Unreplicated Experiment
Calculate se ofHigher Order
EffectsNPP
=> 24
70
Graphical Assessment of Effects
Normally Distributed True Average of Zero Variability Relatively Small
71
Generic First Order Models
# of Variables Model
2 y = b0 + b1X1 + b2X2 + b12X1X2
3 y = b0 + b1X1 + b2X2 + b3X3 + b12X1X2 + … + b123X1X2X3
Where:
b0 = average of all responses bi = Ei/2
72
Model Adequacy Checking
Residual Analysis Normality Assumption
Construct a normal probability plot of the residuals to check departure from normality
Plot of Residuals in Time Sequence Check for positive or negative runs. This implies that the independence
assumption on the errors have been violated Check for change in the error variance over time
Plot of Residuals Versus Fitted Values yij
Check for nonconstant variance Plot of Residuals Versus Other Variables
Check for patterns that might imply that the variable affects the response
^
73
Effect Estimates with 2k Designs
Interaction Main Two Three Four Five Total HigherType Effect Factor Factor Factor Factor Effect Order
Order of 1 2 3 4 5 6 7 8 9 10 Estimates Effects
Interaction (2k - 1) (3rd up)# of Variables
1 1 1 02 2 1 3 03 3 3 1 7 14 4 6 4 1 15 55 5 10 10 5 1 31 166 6 15 20 15 6 1 63 427 7 21 35 35 21 7 1 127 998 8 28 56 70 56 28 8 1 255 2199 9 36 84 126 126 84 36 9 1 511 46610 10 45 120 210 252 210 120 45 10 1 1023 968
We don’t need to expend resourcesto estimate these...
74
Factorial Design Problem
A 23 factorial design problem was employed to study the possible effects of temperature (deg C), time (hr) and stir rate (Rpm) on the output (gallons/hr) of a chemical process.
Run X1 X2 X3 Y
1 20 1 10 28
2 35 1 10 43
3 20 3 10 30
4 35 3 10 47
5 20 1 30 59
6 35 1 30 88
7 20 3 30 65
8 35 3 30 81
75
Popcorn Problem
The owner has developed a new hybrid of seed which may or may not be use the same cooking process recommended for the current seed. One of the processes used by customers to pop corn is the hot oil method, which is addressed in this situation.
The problem is to find the process factors which influence popcorn quality characteristics relative to customer requirements. Characteristics such as the number of unpopped kernels in a batch, the fluffiness or volume of popped corn, the color, the taste, and crispiness are considered.
76
The objective of the experiment is to find the process conditions which optimizes the various quality characteristics to provide improved popping, fluffiness, color, taste, and texture.
The popcorn factors and levels are given in the next slide and use full factorial design to solve the problem.
77
Factor Level 1 Level 2
A: Type of oil Corn oil Peanut oil
B: Amount of oil Low High
C: Amount of heat Low High
D: Preheat No Yes
E: Agitation No Yes
F: Venting No Yes
G: Pan material Aluminum Steel
H: Pan shape Shallow Deep
78
Fractional Factorial Designs
Experimental designs requiring only a fraction of the total number of runs of a full factorial
Always accompanied by some degree of confounding Denoted by: 2k - p
where 2-p = the fraction of the design
k = the number of variables to be investigated
p = number of extra variables to be introduced into a 2k factorial design
2k - p = the number of total runs required
r = resolution of the design
79
Resolution of Fractional Factorial Designs
Resolution Sum of Orders of Confounding DescriptionEffects Confounded
I Main effects non-estimableII 1 + 1 Main effects confounded
with main effectsIII 1 + 2 Main effects confounded with
two-factor interactionsIV 1 + 3 Main effects confounded with
three-factor interactions2 + 2 Two-factor interactions
confounded with 2-factorV 1 + 4 No confounding of main effects
2 + 3 or two-factor interactions witheach other
VI 1 + 5 Greater clarity of effects2 + 43 + 3
80
Exercise
Run X1 X2 X3 X4 Y
1 33 2 0.03 1.9 33.32 39 2 0.03 1.9 40.93 33 4 0.03 1.9 39.44 39 4 0.03 1.9 47.25 33 2 0.09 1.9 28.16 39 2 0.09 1.9 31.17 33 4 0.09 1.9 34.98 39 4 0.09 1.9 419 33 2 0.03 3.5 31.710 39 2 0.03 3.5 41.611 33 4 0.03 3.5 37.612 39 4 0.03 3.5 44.613 33 2 0.09 3.5 28.314 39 2 0.09 3.5 32.215 33 4 0.09 3.5 33.616 39 4 0.09 3.5 39.3
What are the estimatesof all main and interactioneffects?
Which effects are significant?
Write down the predictivemodel and calculate residuals.
Do model adequacy check.
81
Biker problem
Lance Legstrong has one week to fine-tune his bicycle before the early-bird Spring meet. He decides to test seven factors in only 8 runs around the quarter-mile track. It is “saturated” with factors, that is, no more can be added for the given number of runs.
The data for the given problem is given in the next slide.
82
Factor Level 1 Level 2
Seat Up Down
Tires (psi) 40 50
Handle bars Up Down
Helmet Brand Atlas Windy
Gear High Low
Wheelcovers On Off
Generator Off On
83
The time for the ¼ mile track for the 8 runs are given below
Run Time, secs 1 77 2 74 3 82 4 47 5 72 6 77 7 48
8 83
84
DOE is a proactive tool. There is no such thing as a bad experiment - only poorly designed and
executed ones. Not every experiment will produce earth shattering discoveries:
Something will always be learned. New data prompts asking new questions and generates follow-on studies.
The best time to design an experiment is after the previous one is finished. Don’t try to answer all the questions in one study. Rely on a sequence of
studies. Use two-level designs early. Spend less than 25% of the budget on the first experiment. Always verify results in a capping run. Be ready for changes. A final report is a requirement.
90% Planning and 10% doing experiments... Dr. Montgomery
Some Guidelines
85
FINAL NOTE
“When your television set misbehaves, you may discover that a kick in the right place fixes the problem at least temporarily. However, for a long-term solution the reason for the fault must be discovered. In general, problems can be fixed, or they may be solved. Experimental design catalyzes both fixing and solving.”
-- Box, Hunter, and Hunter, “Statistics for Experimenters”, 2nd edition, 2005, John Wiley & Sons, Inc.
86