SIX SIGMA – IMPROVE

---------------------------------------------------------------------------------------------------------------------

1. For a simplex-lattice design the following formula or equation determines:

A. The canonical formula for linear coefficients

B. The portion of each polynomial in the experimental model
C. The canonical coefficients for a cubic model
D. The number of design points

Correct Answer: D

Justification: The last answer is the correct one. The formula is used to determine the
number of design points in a simplex-lattice experiment.

---------------------------------------------------------------------------------------------------------------------

2. Which answer correctly describes the experimental design below?

A. Graeco-Latin Square
B. Latin Square
C. Randomized Complete Block
D. Youden Square

Correct Answer: A

1 www.pmtutor.org Powered by POeT Solvers Limited.

Justification: The Latin Square design has each treatment appearing once and only
once in each row and column. The Randomized Complete Block design place
treatments randomly. The Youden Square design is an incomplete Latin Square. The
design shown is a Graeco-Latin Square (also called Graeco-Roman).

---------------------------------------------------------------------------------------------------------------------

3. An experiment was conducted on five different extruders of polymer, with

readings from 4 die positions, marked A, B, C, and D. Each die position is unique
for each extruder. There were 4 replications per die position. Each extruder had
its own die, and the 4 positions were marked for testing. This is a type of:
A. 5 × 4 factorial experiment
B. Nested experiment
C. Completely randomized experiment
D. Fractional factorial experiment

Correct Answer: B

Justification: This design is not completely randomized and thus does not have all
trials fully randomized (such as the die positions). It is not a true fractional factorial
experiment. Since the die positions are not randomized during the experiment. This is a
nested experiment.

---------------------------------------------------------------------------------------------------------------------

4. A hyper-Graeco-Latin (4 × 4) design is constructed as follows:

Carburetor Type

Car I II III IV

1 AαMϕ BβNχ CγOψ DδPΩ

2 BδNΩ AγMψ DβP× CαOϕ

3 CβOχ DαPϕ AδMΩ BγNψ

4 DγPψ CδOΩ BαNϕ AβM×

2 www.pmtutor.org Powered by POeT Solvers Limited.

 Drivers Tires

A, B, C, D M, N, O, P

Days Speeds
α,β,γ,δ ϕ,χ,ψ,Ω

Assume that car mileage is the output factor. If the above design were converted
to a full factorial design, how many tests would be required for a full factorial?
A. 256
B. 1024
C. 1296
D. 4096

Correct Answer: D

Justification: There are 6 factors at 4 levels. The total number of required tests would
be:

46 = 4096

---------------------------------------------------------------------------------------------------------------------

5. The following is an example of what type of response surface?

A. Rising ridge
B. Maximum or minimum
C. Stationary ridge
D. Minimax

Correct Answer: A

3 www.pmtutor.org Powered by POeT Solvers Limited.

Justification: The drawing is a two dimension example of a rising ridge response
surface.

---------------------------------------------------------------------------------------------------------------------

6. A fractional factorial design for 5 factors (A, B, C, D and E) at 2 levels (+, -) with 8
runs has been designed. In this design, the D and E factors were overlaid into a
standard 2 level, 3 factor design. The design matrix has which effects
confounded?

Test A B C D E
1 - - - + +
2 + - - - -
3 - + - - +
4 + + - + -
5 - - + + -
6 + - + - +
7 - + + - -
8 + + + + +

I. ABC is confounded with E

II. AB is confounded with D
III. AC is confounded with E
IV. BC is confounded with D

A. I and II only
B. I and IV only
C. II and III only
D. II and IV only

Correct Answer: C

Justification: The overlaying of factors D and E resulted in the confounding of the AB

interaction with D, and the confounding of AC with E.

---------------------------------------------------------------------------------------------------------------------

4 www.pmtutor.org Powered by POeT Solvers Limited.

 7. When considering EVOP as a statistical tool:
A. A change in the means indicates that we are using the wrong model
B. An external estimate of the experiment error is necessary
C. EVOP may be extended beyond the two level factorial case
D. We are limited to one response variable at a time

Correct Answer: C

Justification: Answers A and B are filler. Refer to the EVOP diagram below:

The above drawing depicts EVOP for two input variables and one response variable.
However, EVOP may be extended to three or more input variables. More than one
response may also be measured for EVOP trials.

---------------------------------------------------------------------------------------------------------------------

8. Data from mixture experiments are gathered and analyzed through canonical
polynomials. These polynomials are primarily defined as canonical because:
A. The sum of the proportions must equal one
B. The polynomials are different from regular regression equations
C. The number of terms in the polynomial is (q+m-1)! / m!(q-1)!
D. The terms in the polynomials have simple interpretations

Correct Answer: A

5 www.pmtutor.org Powered by POeT Solvers Limited.

Justification: All of the answers A, B, C, and D are true for canonical mixture
polynomials. The term canonical is used to describe relationships that are correlated.
The equations are correlated because the various proportions must sum to one. Thus,
answer A is the best choice. A change in one polynomial affects the other; there is a
lack of independence.

---------------------------------------------------------------------------------------------------------------------

9. The term collinear refers to:

A. Two linear variables
B. A linear interaction
C. Variables being a linear combination of one another
D. Linear correlation

Correct Answer: C

Justification: Variables that are linear combinations of one another are collinear. This
creates high correlation among the variables making analysis very suspect.

---------------------------------------------------------------------------------------------------------------------

10. Experimental design can be effectively used to:

I. Choose between alternatives
II. Select key factors affecting a response
III. Reduce variability of a process
IV. Control uncontrollable noise factors

A. I only
B. II and III only
C. I, II, and III only
D. I, II, III, and IV

Correct Answer: C

Justification: Designed experimentation can affect the first three items well. It will not
control uncontrollable noise factors.

---------------------------------------------------------------------------------------------------------------------

6 www.pmtutor.org Powered by POeT Solvers Limited.

 11. When performing “one experiment with five repetitions,” what are the six
experiments called?
A. Randomization
B. Replications
C. Planned grouping
D. Sequential

Correct Answer: B

Justification: Repeated trials or replications are often conducted to estimate the pure
trial experimental error so that lack of fit may be evaluated. Randomization frees an
experiment from the environment and other biases. Sequential experiments are
conducted one after another, not all at the same time. Adjustments may be made in the
experimentation based upon knowledge obtained. Almost any DOE contains planned
grouping.

---------------------------------------------------------------------------------------------------------------------

12. A L8 (27) design matrix is shown below. What statements are true of this
experimental design?

Column 1 2 3 4 5 6 7
Number
Run O O X O X X O
Number
1 1 1 1 1 1 1 1
2 1 1 1 2 2 2 2
3 1 2 2 1 1 2 2
4 1 2 2 2 2 1 1
5 2 1 2 2 2 1 2
6 2 1 2 1 1 2 1
7 2 2 1 2 2 2 1
8 2 2 1 1 1 1 2

Note:
O – means ok to use column
X – means use, but involved with interactions from other columns

I. Use seven factors if no interactions are present

II. Use with four factors, which will be confounded with 2-factor interactions
7 www.pmtutor.org Powered by POeT Solvers Limited.
 III. Column 3 is suitable for use in a four-factor experiment
IV. The design will be saturated at seven factors

A. I only
B. II and III only
C. I, II, and IV only
D. I, II, III, and IV

Correct Answer: C

Justification: This is a fractional factorial design with 7 factors at 2 levels. Taguchi

placed some restrictions on this design matrix. One precaution is to use seven factors
only if no interactions are present. The design will be saturated since all of the
information will go into the main effects. Another consideration is that four factor designs
will have confounding with 2-factor interactions. Column 3 is not suitable for use for a
main effect. Column 3 will be designated for an interaction term.

---------------------------------------------------------------------------------------------------------------------

13. A Latin square design is noted for its straight forward analysis of interaction
effects. This statement is:
A. True in every case
B. True sometimes, depending on the size of the square
C. False in every case
D. False except for Graeco-Latin squares

Correct Answer: C

Justification: Both Latin and Graeco – Latin square designs are fractional factorials
which will not allow an analysis of interaction effects. The interactions are confounded
with the results of the main effects.

---------------------------------------------------------------------------------------------------------------------

14. Plackett and Burman designs are used for screening experiments. There are
geometric and non-geometric designs. It has been stated that runs of 12, 20, 24,
28, and 36 runs are non-geometric designs. This is because:
A. The runs are in multiples of 4

8 www.pmtutor.org Powered by POeT Solvers Limited.

 B. The non-geometric design has a 2-factor interactions confounded with main
effects
C. The geometric design runs are in powers of 2
D. A PB design of 12 runs can have 11 factors covered

Correct Answer: C

Justification: Answers B and D are true statements but don’t answer the question.
Answer A is a condition for all Plackett and Burman designs. The ability to distinguish a
non-geometric design is that they are not in powers of 2.

---------------------------------------------------------------------------------------------------------------------

15. The iterative approach to DOE refers to:

A. The use of sequential experimentation
B. Assuring the stability of the process during experimentation
C. Assuring the capability of the measurement system
D. Appropriate estimates of experimental error

Correct Answer: A

Justification: The iterative approach to DOE is the recognition that sequential

experimentation will often yield more satisfactory results that “one big experiment.”
Answers B, C, and D refer to traditional experimental assumptions, that must not be
taken for granted (they should be checked).

---------------------------------------------------------------------------------------------------------------------

16. Highly fractional factorial designs are often used as:

A. Simplex designs
B. Orthogonal designs
C. Screening experiments
D. Mixture experiments

Correct Answer: C

Justification: In screening experiments, highly fractional factorial designs are used to

look for factor main effects only. They are called screening because they try to eliminate
seemingly unimportant factors.

---------------------------------------------------------------------------------------------------------------------

9 www.pmtutor.org Powered by POeT Solvers Limited.

 17. In a full factorial experimental design, factors A and B are both tested at 4 levels.
How many experiments are conducted?
A. 8
B. 9
C. 12
D. 16

Correct Answer: D

Justification: This question requires knowledge of experimental design and an answer

review. In a two factor test at four levels, the number of experiments is 42 = 16.

---------------------------------------------------------------------------------------------------------------------

18. A randomized block experimental design is most like a:

A. Higher – order experiment
B. Steepest ascent/descent experiment
C. Taguchi robustness design
D. A Latin or Graeco- Latin design

Correct Answer: D

Justification: Latin and Graeco-Latin designs are block designs. The other answers
don’t fit the question.

---------------------------------------------------------------------------------------------------------------------

19. When comparing a Box-Behnken design with central composite designs, which
of the following statements are FALSE?
I. Box-Behnken designs require fewer runs for 3 factors than CC designs
II. Box-Behnken designs require fewer runs for 2 and 5 factors than CC
designs
III. Box-Behnken designs require fewer factor levels than all CC designs

A. I and II only
B. I and III only
C. II and III only
D. I, II and III

10 www.pmtutor.org Powered by POeT Solvers Limited.

Correct Answer: C

Justification: Note that a negative question response is requested. Item I is a correct

statement. Item II is not correct for two reasons: (1) The Box-Behnken doesn’t exist for
2 factors and (2) versions of CC designs have fewer runs at 5 levels. Item III isn’t
correct because the CCF design also requires three factor levels – the same as Box-
Behnken.

---------------------------------------------------------------------------------------------------------------------

20. How many of the following are considered response surface methodology
designs?
I. CCC
II. CCCF
III. CCI
IV. Box-Behnken

A. I and IV only
B. I, II and III only
C. I, III and IV only
D. I, II, III, IV

Correct Answer: D

Justification: Items I, II and III are varieties of central composite designs. They and the
Box-Behnken are all response surface methodology designs.

---------------------------------------------------------------------------------------------------------------------

21. A lack of statistical knowledge on the experimenter’s part could result in several
items going wrong including:
I. Confounding of undesired effects and interactions
II. Experimental results corrupted by measurement error
III. Inappropriate ranges of control variables
IV. Misidentified control factors causing distorted results

A. I only

11 www.pmtutor.org Powered by POeT Solvers Limited.

 B. II and III only
C. I, II, and IV only
D. I, II, III, and IV

Correct Answer: D

Justification: Unfortunately, all four items can lead to a bad experiment. Coleman and
Montgomery (1993) list six items on the experimenter’s part that can lead to bad results.
The two missing items are: misunderstanding of the nature of interactions, and lack of
appreciation of different levels of error.

---------------------------------------------------------------------------------------------------------------------

22. For a mixture experiment, the design is expressed as {4, 5}. This implies:
A. 5 trials of 4 components
B. 4 proportions, 5 components
C. 4 components, 6 proportions
D. 4 sided cube, 5 points on the cube face

Correct Answer: C

Justification: The {4, 5} design implies 4 components and 6 proportions for the
mixture. Note that the number of proportions equals m + 1, to include both 0 and 1.

---------------------------------------------------------------------------------------------------------------------

23. Box-Wilson central composite designs (CCC and CCI) are rotatable designs.
This implies:
A. The points of the exterior star design can be moved around easily
B. Star points are +/1 unit away from the center of the design space
C. There is a consistent and stable variance about any star point
D. The star points are 1.414 units away from the center

Correct Answer: C

Justification: Answers B and D are fabricated distractors. Answer A is almost correct,

but not quite so. They are rotatable designs because the variance is the same at all
points that are the same distance from the design center. The variance will remain
unchanged when the design is rotated around the center.

---------------------------------------------------------------------------------------------------------------------

12 www.pmtutor.org Powered by POeT Solvers Limited.

 24. The market value of a house is to be determined by 2 input factors. The input
factors are the square feet of living space and the number of bedrooms. If the
input factors are highly correlated with each other, this could indicate that:
I. A multicollinearity condition exists
II. The variables are linear combinations of each other
III. This is an intercorrelation of factors
IV. One of the factors is redundant

A. I only
B. I and III only
C. II, III, and IV only
D. I, II and III and IV only

Correct Answer: D

Justification: All 4 answers are correct. A multicollinearity condition exists per

definition. The variables will be linear combinations of each other, with intercorrelation.
Because of the correlation, one of the variables provides the same information and is
redundant.

---------------------------------------------------------------------------------------------------------------------

25. A 3 × 3 Latin Square would have how many error degrees of freedom?
A. Two
B. Three
C. Four
D. Five

Correct Answer: A

Justification: For the 9 tests there will be two degrees of freedom for the error term.

Source of variation Degrees of freedom Df

Treatments p–1 3–1=2
Rows p–1 3–1=2
Columns p–1 3–1=2
Error (p – 2) (p – 1) (3 – 2) (3 – 1) = 1 (2) = 2
total p×p–1 3×3–1=8

13 www.pmtutor.org Powered by POeT Solvers Limited.

---------------------------------------------------------------------------------------------------------------------

26. Plackett and Burman designs are used for screening experiments. They are 2
level designs with run multiples of 4 instead of powers of 2. Certain designs with
runs of 12, 20, 24, 28, and 36 are considered non-geometric, which means that:
A. Each interaction effect is confounded with exactly one main effect
B. They cannot be represented as cubes
C. These are very economical designs
D. They are used for screening experiments

Correct Answer: B

Justification: This question pertains to the issue of what makes a Plackett and Burman
(PB) design non-geometric. The Plackett and Burman designs are used for screening
experiments and are economical, so answers C and D are not unusual. PB geometric
designs have each interaction effect confounded with exactly one main effect. Thus,
answer B remains as correct. They cannot be represented as cubes.

---------------------------------------------------------------------------------------------------------------------

27. Given the Simplex-Lattice matrix below, what would the proper form of the design
variables: q and m?
X1 X2 X3
0 0 1
0.5 0 0.5
1 0 0
0 0.5 0.5
0.5 0.5 0
0 1 0

A. {6, 3}
B. {3, 2}
C. {3, 4}
D. {2, 3}

Correct Answer: B

14 www.pmtutor.org Powered by POeT Solvers Limited.

Justification: For the design matrix, the form should be {q, m}. q is the number of
component or factors in the blend. The m stands for the proportions at 0.5 or 1.0. The
number of equally spaced levels will be m+1, to make the levels at 0, 0.5, 1.

---------------------------------------------------------------------------------------------------------------------

28. A designed experiment is to be conducted with four factors at three levels. The
factors and levels will be randomized and as uniform as possible. This design is
termed:
A. A one-way ANOVA design
B. A completely randomized design
C. A two-way ANOVA design
D. A Latin square

Correct Answer: B

Justification: The one-way ANOVA and two-way ANOVA are methods of analyzing
designs when one or two factors are to be tested. A Latin square is a more restricted
design form. The completely randomized design is the correct answer.

---------------------------------------------------------------------------------------------------------------------

29. Most modern computer programs will perform an analysis of experimental

residuals. What other techniques can be employed?
I. Control Charts
II. Histograms
III. Normal probability plots
IV. Dot plots

A. I and II only
B. II, III and IV only
C. II and IV only
D. I, II, III and IV

Correct Answer: B

Justification: Control charts aren’t applicable for this situation. Normal probability and
dot plots are widely used. Histograms can be used in some cases.

15 www.pmtutor.org Powered by POeT Solvers Limited.

---------------------------------------------------------------------------------------------------------------------

30. A four factor, three level experiment must be conducted. What are the fewest
number of trials possible if all interactions are ignored?
A. 9
B. 18
C. 27
D. 81

Correct Answer: A

Justification: A full factorial design requires (3)4 or 81 experiments. A one-third

fractional factorial design (properly chosen) can have main effects independent of two
factor interactions, and at least 3 independent 3 factor interactions. A one-ninth
fractional factorial experiment can analyze main effects only (but they are confounded
with 2 factor interactions).

---------------------------------------------------------------------------------------------------------------------

31. A fractional factorial experiment is to be conducted. The main effects will not be
confounded with the two factor interactions. But two factor interactions may be
confounded with other two factor interactions. The design resolution is:
A. Resolution II
B. Resolution III
C. Resolution IV
D. Resolution V

Correct Answer: C

Justification: This is a resolution IV design. See the CSSBB Primer definitions on VIII –
9.

---------------------------------------------------------------------------------------------------------------------

32. Taguchi describes noise as a nuisance factor that is difficult, impossible, or

expensive to control. Types of noise include:
I. Outer noise
II. Inner noise
III. Within product noise
IV. Between product noise

16 www.pmtutor.org Powered by POeT Solvers Limited.

 A. I and II only
B. II and III only
C. I, II, and IV only
D. I, II, III, and IV only

Correct Answer: C

Justification: Taguchi describes 3 types of noise: outer noise, inner noise, and
between product noise. Outer noise is variation in operating environments and human
errors. Inner noise is the aging of the machine, deterioration, and tolerances. Between
product noise is manufacturing imperfection. There is no description for within product
noise.

---------------------------------------------------------------------------------------------------------------------

33. To state that a model in an experimental design is fixed indicates that:

A. The levels used for each factor are the only ones of interest
B. The levels were chosen from a fixed population
C. The equipment from which the data are collected must not be moved
D. The factors under consideration are qualitative

Correct Answer: A

Justification: Answers B, C and D are all filler answers. In fact, if you selected C as
your choice, seriously consider taking the CSSBB exam at a later date. Experimental
design levels are established (or fixed) based on the best advice of people
knowledgeable of the process. A balance design is then considered only at those levels.
Based upon analysis, factors may then be adjusted to other fixed levels for subsequent
experimentation. The objective is to achieve optimum performance.

---------------------------------------------------------------------------------------------------------------------

34. Residuals are estimates of experimental error obtained by subtracting the

predicted responses from the observed responses. The residuals should be
checked for behavior. This can be accomplished by which of the following
methods?
A. Applying a statistical test such as a t test
B. Using a normal probability plot
C. Using a box-and-whiskers box chart
D. Comparing the Xs to the Ys

17 www.pmtutor.org Powered by POeT Solvers Limited.

Correct Answer: B

Justification: Answers A and C are not generally used to check residuals. Answer D is
definitely not the correct choice. A normal probability plot of the residuals can be used to
check the normality of the residuals. In addition, histograms and dot plots are commonly
used.

---------------------------------------------------------------------------------------------------------------------

35. The following Latin Square design of gas mileage by 5 drivers (A, B, C, D, & E)
with 5 different carburetors. The proposed Latin Square design lacks what
requirement?
Carburetor Type

Car I II III IV V

1 A B C D E

2 B A D E A

3 C D A A B

4 D E A B C
5 E A B C D

A. The car versus carburetor interaction

B. A balanced model involving all factors
C. The car and driver interaction
D. The inner matrix of Greek letters

Correct Answer: B

Justification: No interactions among factors are possible with Latin square designs.
The inner matrix of letters does not have to be Greek. A Latin Square must have an
equal number of runs or tests on all factors. Notice that the Drivers do not have an
equal number tested.

---------------------------------------------------------------------------------------------------------------------

18 www.pmtutor.org Powered by POeT Solvers Limited.

 36. As the new experimenter for Six Sigma projects, you have allocated budget
funds for preliminary trial runs. The reason(s) for this would be:
I. Some practice trials are needed, since this has not been done before
II. The plant manager wants evidence that experimentation will work
III. Planning is critical for success
IV. The financial manager wants evidence of possible success

A. I only
B. II and III only
C. I, II, and IV only
D. I, II, III, and IV

Correct Answer: A

Justification: Some of the item options have a grain of truth to them. However, the best
answer is that most experiments are new to plant operating personnel. Some practice
may be needed. As an additional observation, sufficient funds should be available to do
the following: trial runs, actual runs, and verification runs.

---------------------------------------------------------------------------------------------------------------------

37. A simplex design approach is being used to determine the steepest ascent path
for a design that involves three independent variables. An initial experiment
would require how many runs?
A. 3
B. 4
C. 5
D. 6

Correct Answer: B

Justification: Whether using a straight or modified simplex, under the conditions of the
question, one more run, than the number of independent factors, will be required.

Therefore, 3 + 1 = 4.

---------------------------------------------------------------------------------------------------------------------

38. Which of the following is the best description of randomization?

19 www.pmtutor.org Powered by POeT Solvers Limited.

 A. A means of assuring parallel experimentation
B. A technique used to increase the validity of an experiment
C. The repetition of an observation or measurement
D. The relationship between two or more variables

Correct Answer: B

Justification: This is basically a definition question and requires familiarity with

experimental design terms. Answer A is filler. Answer C is a definition or replication or
repeated trials in an experiment. Answer D could represent correlation analysis or a
simple two factor experimental design.

Answer B is the best choice. Randomization enhances the precision and validity or an
experiment by freeing the experiment from biases and the environment.

---------------------------------------------------------------------------------------------------------------------

39. In a mixture experiment, the independent factors are blended into the end
product. All of the following are proper mixtures of three components EXCEPT
for:
A. 0.33, 0.33, 0.33
B. 1.0, 0, 0
C. 0.50, 0, 0.50
D. 0.25, 0.25, 0.25

Correct Answer: D

Justification: Note that a negative response is requested. The blending of the factors is
in proportions from 0 to 1.0. The total must equal 1.0.

---------------------------------------------------------------------------------------------------------------------

40. For the analysis of response surfaces, one special design is a cube with eight
points, supplemented with a 23 factorial. The points are on each axis. These axis
points are at a radius equal to the vertex. There are 15 points (including the
center point). This design is termed:
A. Central Composite design
B. Box-Behnken design
C. Response surface method
D. EVOP

20 www.pmtutor.org Powered by POeT Solvers Limited.

Correct Answer: A

Justification: The type of response surface is what we want to know, so answer C is

not correct. EVOP does not fit the question description. A Box-Behnken design does not
contain an embedded factorial matrix, mentioned in the question. The central composite
design consists of cube with a factorial design of points inside it.

---------------------------------------------------------------------------------------------------------------------

41. If confounding occurs in a three factor experiment (A, B and C) conducted at two
levels, one would expect that factor A would be confounded with:
A. Factor B
B. Factor C
C. The BC interaction
D. Either the AB or AC interaction

Correct Answer: C

Justification: Surely no experimenter would be stupid enough to confound one main

effect with another main effect (Answers A and B are eliminated). For ½ fractional
experiments, the main effect of interest (A in this case) is usually confounded with the
interaction of the other two factors.

---------------------------------------------------------------------------------------------------------------------

42. Many experimental design authors suggest the use of control runs. These control
runs can include tests at the standard process set points. The conduct of a test
with the center points as set points would most clearly imply or suggest:
A. Use of a central composite design
B. Use of a simplex-lattice design
C. Use of a response surface method
D. Use of a full factorial design

Correct Answer: A

Justification: A simplex-lattice design involves a mixture design. A response surface

method is looking at the contours of the response. The full factorial design is a possible
application but is not the most likely answer. A central composite design includes the
center points of the operating process. If the factors are set at the limits of the existing
set points, then the center points are the existing set points.

21 www.pmtutor.org Powered by POeT Solvers Limited.

---------------------------------------------------------------------------------------------------------------------

43. The term level in experimental design refers to:

A. The complexity of the design
B. The specific settings of input factors
C. The number of output responses
D. The number of independent trials or tests

Correct Answer: B

Justification: The term level most often refers to the different settings of an input
factor. The other answer choices are meant to be distractors.

---------------------------------------------------------------------------------------------------------------------

44. Randomized block designs are best suited for:

A. Screening objectives
B. Comparative objectives
C. Response surface objectives
D. Regression model objectives

Correct Answer: B

Justification: Randomized block designs are ideal for comparative experimental testing
of 2 or more factors. Full or fractional factorial designs (including Plackett-Burman) are
best for screening objectives. Response surface objectives require central composite or
Box-Behnken designs.

---------------------------------------------------------------------------------------------------------------------

45. The linear graphs which accompany Taguchi designs have as their objectives:
I. Providing a compact design layout
II. Providing a visualization of the design options
III. Depicting where main factors can be assigned
IV. Depicting where interactions may be evaluated

A. II only
B. I, II and III only
C. II, III and IV only

22 www.pmtutor.org Powered by POeT Solvers Limited.

 D. I, II, III, IV

Correct Answer: C

Justification: The best inclusive answer choice is item II. However, items III and IV are
major components of II and are certainly true statements. Item I is a debatable choice
and should be eliminated in the question writer’s opinion.

---------------------------------------------------------------------------------------------------------------------

46. Experimental design plans usually call for testing a number of factors and
keeping all other conditions as nearly constant as possible. Taguchi Methods call
for similar techniques, but with a twist. Noise conditions are changed a bit to
determine robustness. Where are the noise factors placed?
A. Control array
B. Inner array
C. Outer array
D. Quadratic array

Correct Answer: C

Justification: There are no arrays termed control or quadratic. The inner array consists
of the controllable factors, while the outer array consists of the noise factors. The noise
factors are themselves tweaked to simulate different conditions.

---------------------------------------------------------------------------------------------------------------------

47. In examining the residuals of an experiment, one would expect to see all of the
following results, EXCEPT?
A. Some residuals higher than predicted
B. Some residuals lower than expected
C. Some residuals exhibiting correlation
D. A consistent amount of error across the test range

Correct Answer: C

Justification: Except for answer C, the other answers are correct. Residuals should be
normally and independently distributed with a mean of 0 and a constant variance. There
should not be a correlation of residuals.

---------------------------------------------------------------------------------------------------------------------

23 www.pmtutor.org Powered by POeT Solvers Limited.

 48. If an experimental design objective is to model a response as a mathematical
function of a few continuous factors, then the design objective can be described
as:
A. An optimal fitting of a regression model
B. A screening objective
C. A response surface objective
D. A comparative objective

Correct Answer: A

Justification: For experiments, a comparative objective is to determine if a factor is

significant. A screening objective is to sort out the lesser factors. A response surface
objective is to find the shape of the response geometry. The optimal fitting of a
regression model fits the factors to a response model.

---------------------------------------------------------------------------------------------------------------------

49. The following experimental design model presented below is described as a

second-order model. The model has these characteristics:
I. Curvature is present in the model
II. An interaction of factors is shown
III. Error is present in the model

A. I only
B. II and III only
C. I and III only
D. I, II and III

Correct Answer: D

Justification: All of the above items are present in the design model.

---------------------------------------------------------------------------------------------------------------------

50. A Latin square design is an experimental design which:

A. Cannot be used when estimation of the interaction effects is desired
B. Affords a good estimate of interaction effects
C. May not permit all treatments in every block

24 www.pmtutor.org Powered by POeT Solvers Limited.

 D. May require the need to estimate the parameters during the experimentation

Correct Answer: A

Justification: Answer D is filler. Answer B and C are incorrect statements. A Latin

square design is essentially a fractional factorial experiment which requires less
experimentation to determine main treatment results. However, like other fractional
factorials, the interaction effects are confounded with main effects and cannot be
estimated. One of the assumptions in a Latin Square experiment is that there are no
row, column or factor interactions.

---------------------------------------------------------------------------------------------------------------------

51. An experiment was conducted to determine the effectiveness of a new and

improved feed for hogs. The exact amount (x) of feed was measured and the
change in weight of the hogs become the response variable. If it was determined
that the initial weight of the hogs was correlated to the weight gained, then the
initial weight of the test hog could be termed:
A. Covariate
B. Collinear
C. Confounding
D. Interacting effect

Correct Answer: A

Justification: Two variables are collinear if they are totally correlated. It should be
obvious that there is some correlation of weight gain to initial weight. But this is not the
best answer choice. The initial weight of the hogs could be termed a covariant, since it
is related to weight gain. The heavier the initial weight of hogs, the more weight they
gain.

---------------------------------------------------------------------------------------------------------------------

52. The repeated trials in a designed experiment allow for:

A. First order modeling
B. Determination of experimental error
C. Nested experimentation
D. The resolution of main effects

Correct Answer: B
25 www.pmtutor.org Powered by POeT Solvers Limited.
Justification: Repeated trials allow for determination of experimental error.

---------------------------------------------------------------------------------------------------------------------

53. Any experimental equation that shows two factors multiplied by each other can
indicate:
I. A slope
II. A twist
III. A curve
IV. An interaction

A. I and II only
B. II and III only
C. I and III only
D. II, III and IV only

Correct Answer: D

Justification: Multiplied factors represent interactions, which graphically become

curves or twists in the response surface.

---------------------------------------------------------------------------------------------------------------------

54. A Taguchi L9 design consists of:

I. 9 design runs
II. 3 factor levels
III. 4 input factors

A. I only
B. I and II only
C. II and III only
D. I, II and III

Correct Answer: D

Justification: All of the answers are correct.

26 www.pmtutor.org Powered by POeT Solvers Limited.

---------------------------------------------------------------------------------------------------------------------

55. In a full factorial experiment with 4 factors at 3 levels each, how many trials are
required?
A. 24
B. 12
C. 64
D. 81

Correct Answer: D

Justification: For a full factorial, the number of trials required is equal to the number of
levels raised to the number of factors. For this problem, 3 raised to the 4 th power is 81
trials.

---------------------------------------------------------------------------------------------------------------------

56. An experiment is being run with 8 factors. Two of the factors are temperature and
pressure. The levels for temperature are 25, 50 and 75. The levels for pressure
are 14, 28, 42 and 56. How many degrees of freedom are required for this
experiment to determine the effect of the interaction between temperature and
pressure?
A. 1
B. 2
C. 4
D. 6

Correct Answer: D

Justification: When experimenting, 1 degree of freedom is required to compute the

overall mean, and n-1 degrees of freedom are required for each factor, where n is the
number of levels for the factor. The degrees of freedom required for interactions are the
product of the interactions for the factors involved in the interactions. For this problem,
temperature has 3 levels and pressure has 4 levels. The interaction requires (3 – 1) (4 –
1) = 6 degrees of freedom.

---------------------------------------------------------------------------------------------------------------------

57. Taguchi methodology seeks to reduce the loss to society by:

I. Reducing variation in the process
27 www.pmtutor.org Powered by POeT Solvers Limited.
 II. Reducing variation in the product
III. Utilizing Shingo techniques
IV. Improving the product development process for reduced variation

A. I only
B. II and III only
C. I, II and IV only
D. I, II, III and IV

Correct Answer: C

Justification: All of the above items are suitable Taguchi methods except for item III.
Shingo techniques are the basis of Lean Manufacturing. Variation reduction is the key to
better quality. A better product development process will provide a robust design. It has
been stated that 80% of the manufacturing costs are determined by product design
decisions.

---------------------------------------------------------------------------------------------------------------------

58. The following design equation:

y = b0 + b1x1 + b2x2 + Σ
I. Is a quadratic equation
II. Is a first order equation
III. Accounts for no curvature
IV. Has an interaction

A. I only
B. II and III only
C. I and III only
D. I, II, III, and IV

Correct Answer: B

Justification: The equation does not have any factors raised to the second power or
beyond. Therefore, it has no interactions. Thus, it is a first order equation. The lack of
second order factors also indicates a lack of curvature in the design data.

28 www.pmtutor.org Powered by POeT Solvers Limited.

---------------------------------------------------------------------------------------------------------------------

59. A 32 experiment means that we are considering:

A. Two levels of three factors
B. Two dependent variables and three independent variables
C. Two go/no-go variables and three continuous variables
D. Three levels of two factors

Correct Answer: D

Justification: As a mnemonic device (memory aid) the author uses this: levels are low,
while factors fly. There are three levels of two factors or two factors at three levels. A
total of nine experiments would be conducted.

---------------------------------------------------------------------------------------------------------------------

60. Which of the following design runs are possible for a three factor simplex-lattice
design?
I. 1.333, 0, 1
II. 1, 0, 0
III. 0.5, 0.5, 0.5
IV. 0.333, 0.667, 0

A. I and II only
B. I and III only
C. II and IV only
D. II, III and IV only

Correct Answer: C

Justification: The three mixture components must add to 1.0 (100%). Therefore, only
items II and IV are possibilities.

---------------------------------------------------------------------------------------------------------------------

61. Good planning and execution of an experimental design is essential for success.
Which of the following steps is out of sequence?
1. Select an experimental design

29 www.pmtutor.org Powered by POeT Solvers Limited.

 2. Set objectives
3. Select process variables
4. Execute the design
5. Verify the data
6. Analyze and interpret the results
7. Present the results

A. Select an experimental design

B. Set objectives
C. Verify the data
D. Select process variables

Correct Answer: A

Justification: The NIST (2001) Engineering Statistics Handbook shows that the
selection of an experimental design will be step 3. That is the step that is out of
sequence.

---------------------------------------------------------------------------------------------------------------------

62. From the following experiment yield plot, what can we ascertain was being
tested?

A. A two component full factorial design

B. A two component fractional factorial design
C. A two component mixture design

30 www.pmtutor.org Powered by POeT Solvers Limited.

 D. A two component steepest ascent design

Correct Answer: C

Justification: Note that the scaling for the two components (as a proportion) totals 1.0
in all cases. The plot represents the results of a two component mixture experiment.

---------------------------------------------------------------------------------------------------------------------

63. An incomplete block design may be especially suitable when:

A. There is missing data
B. There is need for fractional replication
C. It may not be possible to apply all treatments in every block
D. There is need to estimate the parameters during the experimentation

Correct Answer: C

Justification: Answers B and D are incorrect statements. An experiment is designed to

collect all necessary data. The design should not be altered based on whether some
data was missing. Thus, answer A is inappropriate. In block designs, some combination
of factors and levels (treatments) may not be possible or desirable. These block designs
may be balanced or partially balanced.

Consider an experiment to be conducted over four days (blocks) evaluating four

competing formulas (treatments). All formulas must be processed through a furnace
which can only accommodate three tests per day. The following balanced incomplete
block design could be used:

BLOCKS TREATMENTS

(DAYS) A B C D

1 X X X

2 X X X

3 X X X

4 X X X

31 www.pmtutor.org Powered by POeT Solvers Limited.

---------------------------------------------------------------------------------------------------------------------

64. EVOP should be used:

A. When there is a manufacturing problem
B. When a process is not in statistical control
C. When an experimenter first begins working on a new product
D. When a process is producing satisfactory material

Correct Answer: D

Justification: Most manufacturing problems should not be addressed by EVOP. EVOP

is used at the end of experimentation when the process essentially has statistical
control. EVOP trials are conducted in the near vicinity of an already satisfactory
process.

---------------------------------------------------------------------------------------------------------------------

65. Which of the following is NOT true in regards to blocking?

I. A block is a dummy factor which doesn’t interact with real factors
II. A blocking factor has 2 levels
III. A block is a subdivision of the experiment
IV. Blocks are used to compensate when production processes restrict
randomization of runs

A. I only
B. II only
C. I, II and III only
D. II and IV only

Correct Answer: B

Justification: Blocking is used to compensate when production processes restrict

randomization of runs. For example, assume an experimental factor is temperature, and
is set at 3 levels. If it is very difficult to move the temperature between the 3 levels, the
experiment could be blocked into 3 groups, low temperature, medium temperature and
high temperature. Each block of trials would be run together. To compensate for not
running the trials in random order, the block is a dummy variable. A dummy variable can

32 www.pmtutor.org Powered by POeT Solvers Limited.

have as many levels as desired. A block can be a subdivision of an experiment. The
question is seeking an incorrect choice, which is item II.

---------------------------------------------------------------------------------------------------------------------

66. In the Taguchi design methodology, what are noise factors?

A. Factors that strongly influence the mean response
B. Factors that impact tolerance design
C. Factors that maximize parameter design
D. Factors that influence variation in the output

Correct Answer: D

Justification: Taguchi referred to signal factors as those factors that strongly influence
the output response and noise factors as those factors that influence variation in the
output response.

---------------------------------------------------------------------------------------------------------------------

67. Central composite designs are one of the classes of response surface designs.
They are popular because:
I. Sequential runs of linear or curvature effects (if needed) can be made
II. They provide much information from a minimum of runs
III. They require only one set of runs for superior results
IV. They are quite flexible over different experimental regions

A. I only
B. II and III only
C. I, II, and IV only
D. I, II, III, and IV

Correct Answer: C

Justification: According to Verseput (2001) the only item not meeting the properties of
the central composite design would be item III. No one claims to have a design that
produces superior results with only one run. All of the other items constitute desirable
properties of a central composite design.

33 www.pmtutor.org Powered by POeT Solvers Limited.

---------------------------------------------------------------------------------------------------------------------

68. If five or more factors are under consideration and the experimenter’s objective is
to construct a design response surface, what is the appropriate first action step?
A. Screen the list down to 2 – 4 factors
B. Select a fractional factorial design
C. Select an appropriate Box-Behnken design
D. Choose a form of central composite design

Correct Answer: A

Justification: Ultimately the experimenter wants to conduct a Box-Behnken or central

composite design. However, there are presently too many factors. First, a screening
design must be conducted.

---------------------------------------------------------------------------------------------------------------------

69. An experiment is to be conducted; you have requested that a statistics professor

develop a design for you. From a practical viewpoint, you should review the
design because:
I. The professor may have limited plant knowledge
II. There may be undesirable combinations of variable control levels
III. The design may cause a violation of known physical laws

A. I only
B. I and II only
C. II and III only
D. I, II and III

Correct Answer: D

Justification: The professor can provide the design but you and the plant personnel will
be conducting the experiment. Montgomery & Coleman (1993) provide a list of items
that the statistician may be lacking. These items include:

1) Unwarranted assumptions of the process,

2) Undesirable combinations of the factors,
3) Violation of known laws of physics,

34 www.pmtutor.org Powered by POeT Solvers Limited.

 4) Too large or small design sizes,
5) Inappropriate confounding,
6) Imprecise measurement,
7) Unacceptable prediction error, and
8) Undesirable run order.

---------------------------------------------------------------------------------------------------------------------

70. When selecting and scaling the process input variables for an experiment, what
is NOT a desirable approach?
A. Include as many important factors as possible
B. Set factor levels at practical or possible levels
C. Combine process measurement responses when possible
D. Be bold, but not foolish, in selecting high and low factor levels

Correct Answer: C

Justification: Note that a negative answer response is requested. Answer C is not a

good idea. If two or more measurement responses are combined, then the important
factor (between them) may remain hidden. The other options (A, B and D) are valid,
prudent actions.

---------------------------------------------------------------------------------------------------------------------

71. An experimental design using a Latin Square of 3 factors and 5 levels will be able
to determine which of the following?
A. Main treatment effects and interactions
B. Interactions only
C. Column and row effects
D. Main treatment effects

Correct Answer: D

Justification: A Latin Square design is a fractional factorial experiment. There are not
expected to be any interactions, since they cannot be measured. Only the main effects
will be determined.

---------------------------------------------------------------------------------------------------------------------

35 www.pmtutor.org Powered by POeT Solvers Limited.

 72. The main objection to designed experimentation in an industrial environment is:
A. Obtaining more information for less cost than can be obtained by traditional
experimentation
B. Getting excessive scrap as a result of choosing factor levels that are too
extreme
C. Verifying that one factor at a time is a most economical way to proceed
D. Obtaining data and then deciding what to do with it

Correct Answer: B

Justification: Note that the key words are main objection, not objective. Answers C and
D are filler. Answer A is the objective of designed experimentation. One main objection
to designed experimentation is the chance of getting excessive scrap because of
extreme factor levels. Other objections are: the chance of getting scrap from
combinations that are a necessary part of the design, machine delays to take samples
and ignorance of the whole experimental process (see the CSSBB Section VIII, cover
quote by Nelson).

---------------------------------------------------------------------------------------------------------------------

73. Which of the following is a correct statement?

A. Variables are confounded if they are difficult to study
B. Two or more variables are confounded if their effects cannot be separated
given the experimental data
C. Variables are confounded if they form a linear combination
D. Two or more variables are confounded if they produce the same effects

Correct Answer: B

Justification: Variables are confounded when the effects of two or more factors are not
separable.

---------------------------------------------------------------------------------------------------------------------

74. A basic L4 Taguchi design is the same as:

A. A two factor, two level, full factorial
B. A two factor, two level, ½ fractional factorial
C. A three factor, two level, full factorial

36 www.pmtutor.org Powered by POeT Solvers Limited.

 D. A test of a single variable at 4 levels

Correct Answer: A

Justification: This is a knowledge based question. The closest answer is choice A. In

application, the L4 is regarded as a two factor, two level design with full treatment of the
interaction. It has the same arrangement, however, as a design that would be used as a
three factor, two level, ½ fractional factorial experiment.

---------------------------------------------------------------------------------------------------------------------

75. Evolutionary Operation (EVOP) is being considered at your plant. You present
these arguments for its use:
I. EVOP is a conservative strategy for improvement
II. Product runs will be conducted with very little scrap
III. For any one series of tests only a few variables are changed
IV. The runs and analysis can be conducted by production operators

A. I only
B. II and III only
C. I, II, and IV only
D. I, II, III, and IV

Correct Answer: D

Justification: EVOP is the sum of all of the above according to Box, Hunter, and
Hunter.

---------------------------------------------------------------------------------------------------------------------

76. If a sample space contains several unknown minimax areas, then what can
happen using steepest ascent methodology?
A. Many tests may be required
B. The yield contours must be ignored
C. The design area around point p must be expanded
D. A wrong answer can result

Correct Answer: D

37 www.pmtutor.org Powered by POeT Solvers Limited.

Justification: If the initial experiment did not detect the possibility of several maximum
peaks (or valleys) then there is a distinct possibility that a lower than optimum result can
occur.

---------------------------------------------------------------------------------------------------------------------

77. A Box-Behnken experimental design is an independent quadratic design. The

endpoints of the design are not used for testing. The Box-Behnken design
possesses which of the following features?
I. It requires 3 levels for each factor
II. It is a rotatable design
III. It always requires fewer runs than other central composite designs

A. I and II only
B. II and III only
C. I and III only
D. I, II and III

Correct Answer: A

Justification: The Box-Behnken designs require 3 levels for each factor and are
rotatable designs. They do not always require fewer runs than other central composite
designs. Central composite designs require fewer treatments when the factors are equal
to or greater than 4.

---------------------------------------------------------------------------------------------------------------------

78. Which of the following DOE definitions is NOT correct?

A. A block is defined as a group of treatments and levels that indicates the total
number of experiments required
B. Replication is defined as the additional experiments needed to increase the
accuracy of a measurement
C. A factor is defined as one of the variables whose influence is being studied in
the experiment
D. Treatments can be defined as the levels assigned to each factor during an
experimental run

Correct Answer: B
38 www.pmtutor.org Powered by POeT Solvers Limited.
Justification: This question requires knowledge of experimental design. A negative
response is requested. All answers are correct with the exception of B. It is true that
taking the average of repeated measurements in metrology will often yield a truer actual
reading (according to the central limit theorem). However, replication in experimental
design is used to estimate the pure trial-to-trial experimental error so that further
analysis of the experimental results can be made.

---------------------------------------------------------------------------------------------------------------------

79. For a full factorial CCC design for four factors, what is the α value?
A. 1.3333
B. 1.6820
C. 2.0000
D. 2.3785

Correct Answer: C

Justification: This question requires a calculation. For a full factorial design, the
formula is:

α = [2k]1/4

Where, k = the number of factors.

α = [24]1/4

α = [16]1/4= 2.0000

For five factors, answer D would be the closest selection. For three factors, answer B is
correct.

---------------------------------------------------------------------------------------------------------------------

80. In the following Simplex-Lattice design, what is the proportion of component X3

at the indicated test location?

39 www.pmtutor.org Powered by POeT Solvers Limited.

 A. 1.000
B. 0.667
C. 0.500
D. 0.333

Correct Answer: D

Justification: The location is the “middle” of the Simplex-Lattice design with ingredients
set at 0.333 for each component, X1, X2 and X3.

---------------------------------------------------------------------------------------------------------------------

81. Sensitivity in experimentation is:

A. Getting the true result
B. Extreme care in data analysis
C. Using the best measuring device
D. Ability to distinguish significant treatment differences in the response variable

Correct Answer: D

Justification: Answer A is the goal of experimentation. Answers B and C reduce the

error of experimentation. Sensitivity is the ability to distinguish differences in the
response (output) variable or variables.

---------------------------------------------------------------------------------------------------------------------

82. Experiments can have many different objectives. Which of the following would be
included in the options?
I. Comparative objective
II. Screening objective
III. Optimized mixture proportions objective

40 www.pmtutor.org Powered by POeT Solvers Limited.

 IV. Response surface determination

A. I, II and III only

B. II and III only
C. I, III and IV only
D. I, II, III and IV

Correct Answer: D

Justification: All of the items (I – IV) are valid goals. They can all be legitimate design
objectives.

---------------------------------------------------------------------------------------------------------------------

83. Taguchi methods use a linear graph to help interpret the corresponding
orthogonal array. For instance, for a L4 array, a linear graph with factors 1 and 2
at the endpoints, and factor 3 at the midpoint indicates:
A. Factor 3 is the interaction of factors 1 and 2
B. That factor 4 is missing, since it is a L4
C. The main factors (1 and 3) are interactions
D. Factor 2 will be the experiment result

Correct Answer: A

Justification: In the interpretation of the linear graph, the endpoints of the graph
indicate the main effects, while the midpoint indicates the resulting interaction term. An
interaction term is possible if the specified column in the design matrix is used.

---------------------------------------------------------------------------------------------------------------------

84. Taguchi designs have as objectives which of the following alternatives?

I. Reduce the quality loss to society
II. Use a development strategy to intentionally reduce variation
III. Identify and develop a parameter that will improve a performance
characteristic
IV. Identify a less expensive design material or method that provides
equivalent or better performance

41 www.pmtutor.org Powered by POeT Solvers Limited.

 A. I, II and III only
B. I, II and IV only
C. II, III and IV only
D. I, II, III and IV

Correct Answer: D

Justification: All of the selections are objectives of Taguchi experimentation.

---------------------------------------------------------------------------------------------------------------------

85. Simplex-lattice mixture designs are characterized by:

I. Regression functions that are referred to as canonical polynomials
II. A requirement there must be the same replicate number of test results for
all test conditions
III. A requirement that mixture proportions total 1.0

A. I and II only
B. I and III only
C. II and III only
D. I, II and III

Correct Answer: B

Justification: Item II is incorrect. In fact, the CSSBB Primer contains an example of an

experiment in which two replicate observations were run at each of the pure blends and
three replicate observations were run at each of the binary blends.

---------------------------------------------------------------------------------------------------------------------

86. When constructing a factorial experiment, which of the following is true?

A. Factorial experiments may not contain any number of levels per factor. They
must be the same for each factor
B. Confounding takes place in factorials when we run a fractional part of the
complete experiment
C. Contrasts and treatment combinations are the same
D. In factorials, the factors must be quantitative

42 www.pmtutor.org Powered by POeT Solvers Limited.

Correct Answer: B

Justification: Answers A, C and D are incorrect statements. Answer B is correct for a

fractional factorial. See how confounding occurs between main effects and interactions
in the diagram below:

A is confounded with BC, B is confounded with AC, C is confounded with AB

---------------------------------------------------------------------------------------------------------------------

87. Information generated in a designed experiment:

A. Always results in an analysis of variance table
B. Is based on the fact that the variance of the sum is the sum of the variances
C. Must always be quantitative
D. May be based on values which are not necessarily numerical

Correct Answer: D

Justification: Answer A is incorrect. An experiment may be analyzed by simple

arithmetic and response table methodology. Answer B is smooth sounding filler.
Attribute information may be generated from an experiment. Questions like: Does the
engine ping? Is the solution cloudy? ; can be addressed.

---------------------------------------------------------------------------------------------------------------------

88. What does a L4 (23) mean?

A. 4 runs to be tested
B. A 2 level, 3 factor experiment at most
C. A total of 32 runs
D. A Latin Square experiment of 8 runs

Correct Answer: A
43 www.pmtutor.org Powered by POeT Solvers Limited.
Justification: Items III and IV are fabricated distractors. L4 implies a Taguchi two-level
orthogonal array. 4 stands for number of runs. It is, at most, a 2 level, 3 factor design.
That is, 3 factors can be used with no interactions, or 2 factors with interactions.

---------------------------------------------------------------------------------------------------------------------

89. The power of efficiency in designed experiments lies in the:

A. Random order of performance
B. The sequential and cyclical procedure of conjecture to design, to analysis and
back to conjecture
C. Hidden replication
D. The large number of possible combinations of factors

Correct Answer: C

Justification: Answer B describes the experimental cycle. Answer A defines how an

experimental plan should be conducted. Answer D is filler relative to this question, but
may be used to describe a complex full factorial design. Answer C is the proper choice.
Even with highly fractional factorial experiments, there is replication of the results of one
factor (against all others) in a balanced design. DOE has a tremendous advantage over
trial and error methods.

---------------------------------------------------------------------------------------------------------------------

90. What form of product design did Taguchi state was the most neglected area?
A. System design
B. Product design
C. Parameter design
D. Tolerance design

Correct Answer: C

Justification: Taguchi refers to system, parameter and tolerance design as design

considerations for products and processes. Of these, Taguchi considered parameter
design to be the most neglected aspect of product design.

---------------------------------------------------------------------------------------------------------------------

44 www.pmtutor.org Powered by POeT Solvers Limited.

 91. A L8 (27) design matrix is shown below. An experimental consisting of 4 factors is
being planned. Interactions are expected in this design plan. Where should the 4
factors be placed?
Column 1 2 3 4 5 6 7
Number
Run O O X O X X O
Number
1 1 1 1 1 1 1 1
2 1 1 1 2 2 2 2
3 1 2 2 1 1 2 2
4 1 2 2 2 2 1 1
5 2 1 2 2 2 1 2
6 2 1 2 1 1 2 1
7 2 2 1 2 2 2 1
8 2 2 1 1 1 1 2

A. Columns 1, 2, 3, and 4
B. Columns 1, 3, 5, and 6
C. Columns 1, 2, 4, and 7
D. Any four columns are acceptable

Correct Answer: C

Justification: Given that there are interactions present, the only possible option is to
place the main effects in columns 1, 2, 4, and 7.

---------------------------------------------------------------------------------------------------------------------

92. Confounding, as it relates to experimental design, means which of the following?

A. Confusing experimental data
B. Mixing interaction and main effects
C. At least 2 repeats of an experimental run
D. That a three factor interaction cannot be measured

Correct Answer: B

Justification: Basic experimental design knowledge and an answer review are

required. Answers A and C are filler. Answer B is correct. Because of a limited number
of experimental tests the effects of main factors and their interactions cannot be

45 www.pmtutor.org Powered by POeT Solvers Limited.

separated. Answer D may be partially correct. However, in this case there could be
confounding of two factor interactions as well.

---------------------------------------------------------------------------------------------------------------------

93. A response surface experiment attempts to accomplish which of the following?

I. Moving toward the optimal point after each experiment
II. Determining the equation of the response surface near the optimal point
III. Inexpensive and simple experimentation, similar to EVOP

A. I only
B. II only
C. I and II only
D. I, II and III

Correct Answer: C

Justification: The response surface method is not like EVOP. It does not promise to be
simple or easy to calculate. Items I and II are correct.

---------------------------------------------------------------------------------------------------------------------

94. The preferred method for determining an optimal point on a response surface is
termed:
A. One factor at a time
B. Steepest ascent
C. The rising ridge approach
D. Mound methodology

Correct Answer: B

Justification: The one factor at a time approach can be used for certain types of
surfaces. It is not adequate for all types. The rising ridge and mound methods are
names for types of surfaces. The preferred method for find an optimal point would be
the steepest ascent approach.

---------------------------------------------------------------------------------------------------------------------

46 www.pmtutor.org Powered by POeT Solvers Limited.

 95. In designed experiments, if the main effect A cannot be distinguished from the
interaction term BC, then it can be stated that:
A. A is a main effect, not significantly different from BC
B. A is diffused into BC
C. A is an alias of BC
D. A and BC have an interaction element

Correct Answer: C

Justification: This requires knowledge of the term alias. An alias occurs when 2 factors
cannot be separated from each other, such as A and BC. All of the other answers are
weak choices.

---------------------------------------------------------------------------------------------------------------------

96. The three assumptions regarding experimental residuals are that:

I. They be normally distributed
II. They be small relative to the main effects
III. They be independently distributed
IV. They demonstrate some constant variance

A. I, II and III only

B. I, II and IV only
C. I, III and IV only
D. II, III and IV only

Correct Answer: C

Justification: The incorrect item choice is item II. Although item II is highly desirable, it
is one of the reasons that people conduct experiments (to find out).

There is an assumption with experiments that the residual or error be normally

distributed, independent and homoscedastic.

---------------------------------------------------------------------------------------------------------------------

97. When considering qualitative and quantitative factors in the same designed
experiment:

47 www.pmtutor.org Powered by POeT Solvers Limited.

 A. The sum of squares for the qualitative factors can still be calculated even
though no numerical scale can be attached to the levels
B. Tables of orthogonal polynomials do not apply because no numerical scale
can be attached to one of the factors
C. The interactions between qualitative and quantitative factors no longer make
sense
D. The tables of orthogonal polynomials apply to both types of factors if the
levels of each are equally spaced

Correct Answer: A

Justification: The author assumes that qualitative and quantitative input factors are
under consideration. Answers B and C are incorrect statements. Answer D is somewhat
flaky because the qualitative factor levels in an experiment are yes or no, Bob or Jane
etc... Answer A is correct. A DOE experiment, which contained several qualitative
factors, was analyzed using response table treatment in the DOE portion of the CSSBB
Primer. The sum of squares of these qualitative factors can be determined by classic
ANOVA table analysis.

---------------------------------------------------------------------------------------------------------------------

98. Experimental design involves process variables including:

I. Input variables
II. Output variables
III. Predictor variables
IV. Dependent variables

A. I and II only
B. II and III only
C. III and IV only
D. I, II, III, and IV

Correct Answer: D

Justification: In experimental design a variety of terms are used to describe virtually

the same thing. Input variables also mean independent variables and predictor
variables. Output variables mean dependent variables.

48 www.pmtutor.org Powered by POeT Solvers Limited.

---------------------------------------------------------------------------------------------------------------------

99. When following the design robustness approach, what step follows parameter
design?
A. Tolerance design
B. System design
C. Concept design
D. Feature design

Correct Answer: A

Justification: The general robustness approach, in order is system (concept) design,

parameter design and tolerance design.

Feature design is meant to be a distractor.

---------------------------------------------------------------------------------------------------------------------

100. What is the major disadvantage of the one factor at a time experimental
strategy?
A. It can be more costly to conduct
B. It fails to account for interactions between factors
C. It offers no ability to vary factors together
D. It is less efficient than designed experiments

Correct Answer: B

Justification: All of the items present good arguments against the one factor at a time
testing. According to Montgomery (2001) the major disadvantage is the failure to
account for interactions.

---------------------------------------------------------------------------------------------------------------------

101. Modern designed experiments can best be defined as having a

methodology to:
I. Vary independent variables simultaneously
II. Perform trials in a carefully planned manner
III. Change factors one or two at a time
IV. Hold many factors constant

49 www.pmtutor.org Powered by POeT Solvers Limited.

 A. I only
B. I and II only
C. II and III only
D. I, II, III, and IV

Correct Answer: B

Justification: Designed experiments are able to vary independent variables (factors)

simultaneously in a carefully planned manner to obtain the results needed.

---------------------------------------------------------------------------------------------------------------------

102. The advantage of using the “modern designed” method of

experimentation, rather than the classical, is that:
A. The procedure by which everything is held constant except the factor under
investigation is less costly
B. Experimental error is recognized but need not be stated in quantitative terms
C. Fewer terms and measurements are needed for valid and useful information
D. The sequence of measurement is often assumed to have no effect and
provides greater test freedom

Correct Answer: C

Justification: The key to this question is the meaning of the word classical. To most
authors, this term refers to varying one factor at a time while holding all other factors
constant. Although this approach may work for very simple problems, it causes havoc
with moderately complex systems. The “fixed” factors do vary which can waste time,
effort and money. This traditional approach can yield invalid or inconclusive results.
Modern design experimentation, (including fractional factorials, improved three factor
designs, and Latin square logic) squeeze a large amount of valid information from a few
trials.

---------------------------------------------------------------------------------------------------------------------

103. Which of the following statements is true about a Latin Square design?
A. It allows for the effects of interaction in the design
B. It does not allow for the effects of interaction in the design
C. It eliminates the need to ensure that the effects of interaction are additive
D. It is useful because the underlying distribution does not need to be normal

50 www.pmtutor.org Powered by POeT Solvers Limited.

Correct Answer: B

Justification: This question requires knowledge of Latin Square experimental designs

and an answer review. Latin square designs are highly fractional factorial experiments
which permit an analysis of main effects only. Interaction effects cannot be determined;
they are confounded with the main effect results.

---------------------------------------------------------------------------------------------------------------------

104. An experiment has 3 factors with 2 levels and 1 factor with 3 levels. What
is the minimum number of trials necessary if all interactions are ignored?
A. 4
B. 5
C. 6
D. 7

Correct Answer: C

Justification: When experimenting, 1 degree of freedom is required to compute the

overall mean, and n-1 degrees of freedom are required for each factor, where n is the
number of levels for the factor. The degrees of freedom required for interactions are the
product of the interactions for the factors involved in the interactions. This experiment
has 1 degree of freedom for the overall mean, 3 factors with 2 levels, which require 1
degree of freedom each, and 1 factor with 3 levels, which requires 2 degrees of
freedom. Thus, 6 degrees of freedom are required, so a minimum of 6 trials are
required.

---------------------------------------------------------------------------------------------------------------------

105. A 1/3 fractional factorial experiment of 9 runs with 3 factors at 3 levels and
interactions can be designated as:
A.
33-1
B. L9
C. 32
D. 1/3(3 × 4)

Correct Answer: A

Justification: Answer B (L9 designation) would apply to an orthogonal array of 9 runs

at 3 levels and 4 factors. But the 3rd and 4th factors must not have any interactions.
51 www.pmtutor.org Powered by POeT Solvers Limited.
Answer C gives 9 runs, but indicates 3 levels and only 2 factors. Answer D is a
distractor. The first answer (A) provides the needed solution. Box, Hunter, & Hunter
(1978) provides the following formula:

---------------------------------------------------------------------------------------------------------------------

106. When attempting to go in the direction of steepest ascent, in a two

independent factor experiment, the best general approach is to:
A. Use a modified simplex model
B. Go perpendicular to the contour lines and adjust as necessary
C. Use the equation generated from the experiment as a roadmap
D. Make sure there are no minimax conditions

Correct Answer: B

Justification: The best general approach is to go perpendicular to the contour lines and
adjust as the contour lines change. Simplex or modified simplex models can be used
but several other alternatives exist. The equation may not be fully developed or
accurate depending upon the stage of experimentation.

---------------------------------------------------------------------------------------------------------------------

107. An experiment with two factors, in which all levels of one variable are run
at each level of the second variable, is called a:
A. One-way experiment
B. Latin square experiment
C. Factorial experiment
D. Fractional factorial experiment

Correct Answer: C

Justification: Depicted below are comparisons of full and fractional factorial designs for
three factors at two levels.

52 www.pmtutor.org Powered by POeT Solvers Limited.

A one-way experiment involves one factor. Latin square experiments are fractional
factorials. Therefore, answer C must be selected.

---------------------------------------------------------------------------------------------------------------------

108. Taguchi methods employ a series of arrays in the experimental design.

Which, of the following choices, includes noise factors?
A. Signal-to-noise arrays
B. Inner arrays
C. Outer arrays
D. Replication arrays

Correct Answer: C

Justification: Answers A and D are distractor choices. The inner array contains the
controllable factors, while the outer array contains the uncontrollable factors, such as
noise factors.

---------------------------------------------------------------------------------------------------------------------

109. Plackett and Burman non-geometric designs contrasted with PB geometric

designs have which flaws?
I. They must be run in multiples of 4
53 www.pmtutor.org Powered by POeT Solvers Limited.
 II. They are not fractional factorials
III. They are only main effects designs
IV. They are screening experiments

A. II and III only

B. I, II, and III only
C. I, III, and IV only
D. I, II, III, and IV

Correct Answer: A

Justification: Plackett and Burman designs are always run in multiples of 4 and are
screening experiments. Answers I and IV are true for every design. Plackett and
Burman non-geometric designs are not fractional factorials. They are main effect
designs only.

---------------------------------------------------------------------------------------------------------------------

110. In every experiment there is experimental error. Which of the following

statements is true?
A. This error is due to lack of uniformity of the material used in the experiment
and to inherent variability in the experimental technique
B. This error can be changed statistically by increasing the degrees of freedom
C. The error can be reduced only by improving the material
D. In a well-designed experiment there is no interaction effect

Correct Answer: A

Justification: Answer D is incorrect because many experiments are designed to

measure interactions. Answer C is wrong because error is often inherent in other areas
of the experimental technique (measurement for example). Answer B is off target
because only a more refined estimate of error can be determined by increasing the
degrees of freedom.

---------------------------------------------------------------------------------------------------------------------

111. A designed experiment of three factors (A, B, & C) at two levels was
conducted. The eight runs were analyzed, suggesting that one level of factor A

54 www.pmtutor.org Powered by POeT Solvers Limited.

 showed significant improvement. The Plant Manager stated that no additional runs
are needed. Your response is:

A. The experiment did its job and should be closed

B. The results of the test should be implemented immediately
C. Additional replications are needed to verify the experiment
D. An EVOP should be conducted to improve the process further

Correct Answer: C

Justification: Running an EVOP is not appropriate at this time. The process may not
be stable yet. The experiment itself may require additional replications to verify the
results. Experimental budgets should be sufficient to allow for addition runs.

---------------------------------------------------------------------------------------------------------------------

112. Which of the following purposes are served by replicating an experiment?

I. Provide a means for estimating the experimental error
II. Increase the number of treatment effects
III. Improve the precision of estimates of treatments effects

A. I and II only
B. I and III only
C. II and III only
D. I, II, and III

Correct Answer: B

Justification: Item II is incorrect. Replicating an experiment can easily be done without

increasing the number of treatment effects. By extension, the only correct answer must
be B. Replicating an experiment permits the estimation experimental error and improves
the precision of an F test for significance of treatments. Refer to the example of an A ×
B factorial experiment in the CSSBB Primer, Section VIII where three replications were
conducted.

---------------------------------------------------------------------------------------------------------------------

55 www.pmtutor.org Powered by POeT Solvers Limited.

 113. A prime reason to consider the use of Evolutionary Operation (EVOP) in
industrial facilities would be:
A. When a bold strategy for improvement is considered
B. When existing industrial processes are not running at their highest efficiencies
C. To test large samples in a conservative manner
D. Because calculations for EVOP are simplified and will lead to better results

Correct Answer: B

Justification: Answer A is incorrect; other test methods are bolder than EVOP. Large
test samples are carried out in a conservative nature, but that is why EVOP is
performed. The calculations in EVOP are simplified, but may not lead to better results.
Box, Hunter, & Hunter (1978) states that answer B is the best reason for EVOP.

---------------------------------------------------------------------------------------------------------------------

114. The smallest run number possible to examine the main effects of 22
factors at 2 levels is:
A. 23
B. 24
C. 44
D. 56

Correct Answer: B

Justification: Since the only interest is in main affects, a screening experiment is

warranted. The most economical choice is a form of Plackett – Burman design. In these
designs, the number of run must be divisible by 4 and there must be at least one more
run than there are factors.

In this case, 22 factors require 24 tests. 21 or 23 factors would also require 24 tests.
The question writer would throw in another “fake” factor such as the experimenter
standing on one foot versus two feet and hope that it doesn’t show up an important
factor in the analysis.

---------------------------------------------------------------------------------------------------------------------

115. Which of the following pure mixture design statements is false?

A. The major design assumptions are similar to those of factorial designs
B. Three or more ingredients constitute the entire sample space
56 www.pmtutor.org Powered by POeT Solvers Limited.
 C. Mixture components, as a proportions, must sum to unity
D. The measured response is dependent on ingredient proportions and not on
mixture amount

Correct Answer: B

Justification: Note that a negative response is requested. In mixture experiments, the

ingredients total to unity (1) and the ingredient proportions are more important than
amount. Mixture amount would be considered only if a composite mixture and process
variable experiment is being conducted.

The assumption of independence, normalcy, and homoscedasticity is similar to that for

factorial designs. Answer B is incorrect because there can be only two components in
the mixture.

---------------------------------------------------------------------------------------------------------------------

116. An experiment yielding the following equation:

What cannot be said about the response surface?

A. It contains more than a 3D picture
B. It contains a curve component
C. It contains two slopes
D. It contains a twist component

Correct Answer: A

Justification: Note that the question is looking for a negative response. Answer B is

correct. is a curve component. Answer C is correct. 12.7 X1 and 16.1 X2 are

slopes. Answer D is correct. 6.4 X1X2 represent a twist. The dimension of the response
is in 3D, representing yield for components X1 and X2.

57 www.pmtutor.org Powered by POeT Solvers Limited.