MTRL 460 MOMP

MTRL 460 MOMP Tutorial 3 (20%): T3A(3%)+T3B(6%)+T3C(11%)

Tutorial 3A: Introduction to Designed Experiments - Experiment on Yourself
The three parts of the Tutorial 3 [20%] include T3A (3%), T3B (6%), and T3C (11%) focus on various aspects of Design of
Experiments. In the 3 parts of Tut3A (i,ii,iii) student teams will experiment using full 2-level Factorial Designs with 2
variables (2FD2) and 3 variables (2FD3), and self-study materials (Tut3Aiii) in preparation for the remaining parts B,C of
this Tut3. Tut3B,C use a custom-made computer simulation program LAB, and commercial statistical software DX, to
design and analyze Fractional Factorials (3B) and Response Surfaces Methods RSM (3C) for a specific Process
Optimization task assigned to any given Team – see the separate description of Tut3B,C. The report for all three Parts
of the Tutorial 3A,B,C, in the form of PDF File named "MTRL460-2021-TutRep3-Team#.pdf", is due Mon Dec 13,
11:59pm. (PLEASE include the file name in the Subject Line of your mail). The Reports score max 85% for fulfilling all
the expectations; the balance of 15% is discretionary to the marker, for exceeding the expectations.
Factorial Design of Experiments (DX) in a Nutshell
Experiments should be optimally "designed" to generate the maximum information in minimum number of
experiments. In this course we focus on the 2-level "Factorial Designs" with k-variables 2FDk, where k=2 in the first
part (i) and k=3 in the second part (ii) of this tutorial Tut3A. Later in Tut3B,C we will also explore RSM.

2FDk experiments run at 2 levels (LOW and HIGH level, often coded as -1 and +1) and require 2k experiments to
cover all combinations of the "low" and "high" levels of the k variables. For example, the "friction coefficient" 2FD2
experiment in Tut2 probed the effects of k=2 variables in 22 =4 experiments, testing the surface roughness at two
levels (-1=LOW, +1=HIGH), and the contact area (-1=LOW, +1=HIGH) according to the following Design Matrix
(mean data from TT are used again here for the illustration only):
Exp. # Roughness Contact Area Exp result fm
factor factor ("response")
1 LOW (-) LOW (-) 0.448
2 HIGH (+) LOW (-) 0.687
3 LOW (-) HIGH (+) 0.536
4 HIGH (+) HIGH (+) 0.576
Average (mean) of all experiments: 0.562

The effect of any factor is calculated by taking average response for that factor at a HIGH (+1) level, and subtracting
the average response for that factor at a LOW (-1) level. For example, using the numbers above, the effect of
surface roughness =0.140 (i.e. it is relatively high, ~25% of the average of all experiments=0.562). Similarly, the effect
of surface area = 0.012 (i.e. relatively low, only 2% of the average).

One more important effect can be calculated from these data: it is the effect of interaction of the roughness and the
contact area variables. The lack of interaction means that the effect of e.g. roughness is the same no matter if the second
variable is high or low. Conversely, strong interaction shows when the effect of e.g. roughness is very different when the
second variable i.e. contact area is at high level, then when it is at low level. The ability to detect/quantify interactions is
unique for Designed Experiments, and absent in other methods, e.g. one-at-the-time experimentation.

The effect of interactions is calculated by taking average value of the responses when the variables are all at low or all at
high levels, and subtracting the average value of the responses when the variables are at mixed levels (LOW-HIGH) –
see the fig below. The "friction coefficient" DX example is illustrated below for the calculation of the effect of surface
roughness and surface contact area interaction X1X2, which
is 0.100. It is large relative to the average of absolute values
of all the responses, ie ~17%, as opposed to the relatively
small effect of the surface contact area (2%). We can safely
assume that the 2% effect is due to experimental error, and in
fact it should be zero (according to the theory of friction)

In this Tut3A we will use similar design and analysis of 2FDk

experiments with (i) k=2 variables (in 22 =4 experiments) and
(ii) k=3 variables (in 23 =8 experiments).

The design (i) is run and analyzed exactly as the above

"friction coefficient" example. The design (ii) will result in
several more interaction effects, between the pairs of 2
variables (three interaction effects) and one effect of the
interaction of all 3 variables.

Page 1
 MTRL 460 MOMP
Tut3A(i): 2FD2 Experiment “BLIND BALANCE”
Two-level Factorial Design with 2 Variables
1. OBJECTIVE: to determine the effects of the support surface and body rotations, on body "blind balance" abilities.

2. EXPERIMENTAL PLAN: Design 2FD2 with the following two variables and one response:

2 Variables: X1 Surface type; X1 (low level {-}): floor; X1 (high level {+}): foam
X2 Rotations; X2 (low level {-}): 0 rotations 0R; X2 (high level {+}): 1 rotation 1R

1 Response: Y: In-balance time (seconds)

DESCRIPTION: Our ability to stand on one leg seems to be linked to our health – see the short article attached at end of
this Tut2A description by Prof D. Skelton. In our Designed Experiment we attempt to assess students’ ability to stand on
one leg with closed eyes (expressed as time [s] before touching the floor due to loss of balance), on two types of support:
1. Regular (hard) floor, and 2. Soft support, on polymer foam pad. Additionally, we attempt to assess the effect of mild
disturbance of our sense of balance by repeating the tests after one full rotation of body.

Here is what to do: Close (cover) your eyes. Count how many seconds can you stand with one leg up some 10-20 cm
above the floor, (1) (-) on the floor directly vs (+) on the foam pad. (2) Repeat the same set of experiments immediately
after one full rotation of your body with closed eyes. Analyze the experiment following the above "friction coefficient" DX
example, i.e. determine the variables effects and interactions. NOTE: do not attempt this experiment if you feel
uncertain/concerned about your balance holding abilities or any other safety issues–use TT’s data instead

3. DATA COLLECTION. In order to simultaneously test both variables X1, X2 at both levels (-) and (+), the following 22=4
“factorial” experiments are run and responses column Y filled in. Take the average of Y from at least three repetitions A,
B, C, and write in your data; {-} and {+} represent the typical method of coding low and high level of variables in DOE; TT#
represent data collected on himself. The resulting Data Entry Table looks as follows:

Your Data TTs Data [sec]

Exp# X1 [surface] X2 [rotations] X1 .X2 Y[sec]A Y[sec]B Y[sec]C Y[sec]Ave
1 floor {-} 0 Rot {-} {+} TT 18.7
2 foam {+} 0 Rot {-} {-} TT 6.3
3 floor {-} 1 Rot {+} {-} TT 6.7
4 foam {+} 1 Rot {+} {+} TT 5.7
Ave All: Ave All TTs: 9.4s

4. DATA ANALYSIS. The result of such 2FD2, i.e. 2-level factorial design with 2 variables (in 22 = 4 experiments) can be
conveniently plotted on a “square plot”, with the response values Y, i.e. average balance time, written at corners
(compare with your own numbers-could be very different – cross TTs numbers and write yours):

(+) 6.3
5.7

X1

(-)
18.7 6.7
X2
(-) (+)

The total “effect” L of any given variable on the experiment outcome is calculated as the difference between all the
average responses when the variable is at high and low levels. Therefore, the effect of the surface support (foam vs
floor) LX1 is calculated for TT as:

Page 2
 MTRL 460 MOMP

LX1 = (6.3 + 5.7)/2 – (18.7 + 6.7)/2 = -6.7 s = -69% of the average: Strong Negative Effect of the Surface Type

Similarly the effect of the body rotation (1 Rot vs 0 Rot ) LX2 is calculated as:

LX2 = (6.7 + 5.7 – 18.7 – 6.3)/2 = -6.3 s = -67% of the average: Strong Negative Effect of the Rotations

As expected there is an overall strong negative effect of both variables on the response, meaning that the response Y
decreases when we move from the low to high level of both variables X1, X2. YOUR DATA WILL BE DIFFERENT!

The effect of the interaction is calculated by taking the average difference between (1) the effects when both variables
are at high or both at low level, and (2) the effects when all the variables are at mixed levels (i.e. one at high level and
another at low level). For the current experiment with TT’s data included in the above Data Entry Table the effect of the
variables interaction is therefore calculated as follows (compare with your own numbers):

LX1X2 = (18.7 + 5.7 – 6.7 – 6.3)/2 = 5.7 s = 61% of average: Strong Interaction Effect

The effects calculation is simplified if we use the “coded” levels of variables, i.e. +1 for the high level and -1 for the low
level. Then, the effect of any given variable is obtained by multiplying the column of coded levels by the column of the
responses in the Data Entry Table, and dividing by the number of data points at each level (here = 2) For example, for
variable X1 (surface type) the results are the following:

LX1 = (-1*18.7 + 1*6.3 -1*6.7 +1* 5.7)/2 = -6.7 s, i.e. identical as calculated before.

The same is valid for calculation of the interaction effect, wherein the column of “interaction levels” is obtained by
multiplying the coded levels of the variables – see the column X1X2 above, resulting in:

LX1X2 = (+1*18.7 - 1*6.3 -1*6.7 +1*5.7)/2 = 5.7 s, i.e. identical as calculated before.

Compilation of your data:

The effect of X1 surface: L1 = ……………

The effect of X2 rotation: L2 = ……………

The effect of interaction of X1 surface and X2 rotation: L12 = ……………

CONCLUSIONS FROM TT's EXPERIMENT: Both variables have strong negative effect on the response; both variables
also interact strongly, at least for TT. The effect of soft-support (foam) on the floor is so strong on his balance that rotation
with foam is not much different than rotation without the foam.

CONCLUSIONS FROM YOUR EXPERIMENT (for your data): {{write your conclusions for at least one member of
the team performing the experiment; comparison between different members will score additional points}}

Page 3
 MTRL 460 MOMP

Tut3A(ii): 2FD3 Experiment “BLIND COORDINATION”

Two-level Factorial Design with 3 Variables
1. OBJECTIVE: to determine the effects of distance (Close, Far), hand (L,R) and body rotations ( No Rotation, One
Rotation), on your body "blind coordination" abilities.

2. EXPERIMENTAL PLAN: 2FD3 with the following three variables and one response:

Variables: X1 hand-to-target distance; X1 low level {-}: close X1 high level {+): far
X2 hand X2 low level {-}: right X2 high level {+}: left
X3 Rotations X3 low level {-}: 0 rot 0R X3 high level {+}: 1 rot 1R
“Response” Y: shots on-target (as % success of all shots)

DESCRIPTION: Attach with the Scotch tape 2 pages of white paper with ~10 cm diameter circles drawn: green circle on
your right and red circle on your left. Sit at the table, without body rotation (X3{-}) , having green pen in your right
hand (X2{-}). Close (cover) your eyes, and aim at the circle on your right (i.e. attempt to make a dot inside the circle)
(X1{-}), then on your left (X1{+}) , then right again, until 10 “shots” are attempted with the green pen at each circle. Then
attach a new set of targets to the table and repeat the same after one full rotation of your body with eyes covered (X3{+}).

Then take red pen to your left hand (X2{+}) and repeat the above plan for 10 shots at each circle, first without rotations
(X3{-}), and then attach new set of targets to the table and repeat the same after one full rotation of your body with
eyes covered (X3{+}).

Calculate the % success rate (“the response”) of the shots within the target circle; assume the shot on/touching the rim
being on target. Perform ~3 repetitions, time permitting, to obtain the average % success rate. Plot the data in the “cube
plot” (see example on the next page) and analyze this 2FD3 Factorial Experiment: determine the variables effects and
their interactions, following the example below for TTs data. NOTE: do not attempt this experiment if you feel
uncertain/concerned about your coordination abilities or any other safety issues – use TT’s data instead

3. DATA COLLECTION. In order to simultaneously test all the three variables at both (-) and (+) levels the following 2 3=8
“factorial” 2FD3 experiment were run and the responses column filled in (TT’s # represent data collected on himself). The
“TT’s Effects” of each variable and the interactions, seen in the last row, were calculated as the difference between all
responses when the variable is at high and low levels, or by multiplying the responses Y column by the respective
variable column):

Main Interactions
Factors Your TT's
Distance Hand Rotation Data Data
Exp # X1 X2 X3 X1X2 X1X3 X2X3 X1X2X3 Y(%) YTT (%)
1 -1 -1 -1 1 1 1 -1 70
2 1 -1 -1 -1 -1 1 1 50
3 -1 1 -1 -1 1 -1 1 40
4 1 1 -1 1 -1 -1 -1 30
5 -1 -1 1 1 -1 -1 1 40
6 1 -1 1 -1 1 -1 -1 30
7 -1 1 1 -1 -1 1 -1 50
8 1 1 1 1 1 1 1 20
TT's Effects: -70/4 -50/4 -50/4 -10/4 -10/4 +50/4 -30/4 41=average

4. DATA ANALYSIS. The results of such 2FD3, with 3-variables at 2 levels (23=8 experiments) factorial design of
experiments can be conveniently illustrated on a “cube plot”, with the response values (i.e. average % of the successful
hits) at the cube corners – see below. The total “effect” L of any given variable on the experiment outcome is calculated
as the difference between all responses when the variable is at high and low levels, or by multiplying the above
coded levels of X1 by the responses column Y. Therefore the effect of the hand distance (far vs close) LX1 is calculated
for TT (compare with your own numbers-could be very different – cross TTs numbers and write yours) as follows:

L1 = (-Y1 + Y2 - Y3 +Y4 - Y5 + Y6 - Y7 + Y8 )/4 = (-70+50 -40+30 -40+30 -50+20)/4= -17.5 = 43% of the average

Page 4
 MTRL 460 MOMP

Likewise, the other two main effects are negative: L2 = -12.5 , and L3 = - 12.5 . It appears therefore that for TT the
distance has higher negative effect than hand (left or right) and body rotation. The interaction effects are more difficult to
illustrate on the cube plot, but are simply calculated by multiplying the columns of the interactions by the column of the
responses, i.e. the interaction of distance and hand L12 is:

L12 = (+Y1 - Y2 - Y3 + Y4 + Y5 - Y6 - Y7 + Y8 )/4 = (+70-50-40+30+40-30-50+20)/4 = -2.5 = 6% of average

The other two-variable interactions are L13 = -2.5 , and L23 = +12.5, and the effect of interaction of all three variables is

L123 = (-Y1 + Y2 + Y3 - Y4 + Y5 - Y6 - Y7 + Y8 )/4 = ( -70+50+40-30+40-30-50+20)/4 = -7.5 = 18% of average

Here compile YOUR data according to the examples above

Compilation of your data:

The effect of X1 distance: L1 = ……………

The effect of X2 hand: L2 = ……………

The effect of X3 rotation: L3 = ……………

The effect of interaction of X1 distance and X2 hand: L12 = ……………

The effect of interaction of X1 distance and X3 rotation: L13 = ……………

The effect of interaction of X2 hand and X3 rotation: L23 = ……………

The effect of interaction of X1 distance and X2 hand and X3 rotation: L123 = ……………

CONCLUSIONS FROM THE EXPERIMENT (for TT’s data): The variable X1 (the distance of his hand from the target)
has the strongest negative effect; however, the variables X2 (hand) and rotation (X3) appear to also have significant
effects on the response and both these variables also interact strongly, at least for TT. Other interactions are relatively
weaker. Further interpretation of these data could proof difficult and speculative, possibly involving neurons-muscle
interactions as a function of disturbance level ("rotations") and hand reach-out distance.

CONCLUSIONS FROM THE EXPERIMENT for your data: {{write your conclusion for at least one member of the
team performing the experiment}}

Page 5
 MTRL 460 MOMP
Tut3A(iii): Self-Study Short Courses and Tutorials on DOE, provided by StatEase
OBJECTIVE: to self-study the Seven Sections 1-7 of the materials provided by Design Expert (DX) software authors
(StatEase), in preparation for DX use in Tut3B,C of MTRL460. This Tut3A(iii) does not require any report, and may
be performed in-class and/or out of class, in-teams or individually.

DESCRIPTION: The first four Sections 1-4 of the self-study Tutorial 3A(iii) are to be found on the StatEase website at
https://www.statease.com/training/stat-ease-academy.html ; you will need to create an account and “order” the courses,
but there is no payment required. These include the following four short (~1 hr each) courses, as described by StatEase:

1. PreDOE: Basic Statistics for Experimenters

PreDOE is a hands-on presentation of practical statistical tools needed to succeed using design of experiments. Through
the use of interactive exercises and tutorials, you will learn concepts of statistical thinking, risk management, t-testing and
analysis of variance (ANOVA).

2. Four Easy Steps to Effective Factorial Design

Learn the four steps used to plan effective design of experiments (DOE). Proper planning is key to a successful DOE.
This course is intended for anyone who would like to do DOE, but doesn't know where to start. For example, an engineer
who wants to improve the quality of a product or reduce costs, or a chemist who wants to increase the yield of a chemical
process. Any product or process that has controllable inputs (factors) and measurable outputs (responses) can be
improved by using DOE. This course will teach you how to build a good factorial DOE.

3. Finding the Vital Settings via Factorial Analysis

Learn how to analyze a two-level factorial experiment. Designed as a follow-up to the above 2. "Four Easy Steps to
Effective Factorial Design" course, you will analyze the DOE studied there. Additional case studies are used to reinforce
the analysis techniques. By the end of the course you will know how to fully analyze a factorial DOE including:
–Selecting the effects for modeling
–Interpreting an analysis of variance (ANOVA)
–Using diagnostics to ensure a valid model
–Interpreting graphs to optimize your product or process.

4. Fractional factorial DOE e-learning: How to Save Runs with Fractional Factorial Designs
Learn how to save experimental effort by using fractional factorial designs. Designed as a follow-up to the "Finding the
Vital Settings via Factorial Analysis" web-based e-learning course, you will learn about aliasing and how to choose the
right fractional factorial to meet your DOE objectives. Explore the screening, characterization and optimization strategy to
find optimum factor settings for your product or process.

The remaining three Sections of the self-study Tutorial 3A(iii) are in 460-ARA3B2021-DXTutorials (posted on Canvas)
and include the following Sections 5, 6, 7 ; each of these is a step-by-step illustration of the StatEase DX software,
wherein students will follow the instructions and execute them using DX software (to execute you need to download a
limited-time free copy from https://www.statease.com/software/dx-trial.html ; please also note: the recent 3 versions of
DX by StatEase DX11, 12, 13, are identical in terms of statistical analysis and very similar in terms of the capabilities,
graphical interface etc):

5. Two-Level Factorial: pages 1-38 of ARA3B

This tutorial demonstrates the use of Design-Expert software for two-level factorial designs. These designs will help you
screen many factors to discover the vital few, and how they interact

6. Response Surface – The Basics: pages 39-71 of ARA3B

This tutorial shows how to use Design-Expert software for response surface methodology (RSM). This class of designs is
aimed at process optimization. A case study provides a real-life feel to the exercise.

7. Response Surface - Optimization: pages 72-94 of ARA3B

This tutorial shows how to use Design-Expert software for process optimization and builds on the skills gained in the
above Section 6 RSM-The Basics.

Page 6
Academic rigour, journalistic flair

Standing on one leg regularly is good for you.

Storytimestudio/Shutterstock

Standing on one leg is a sign of good health – and practising

is good for you too
October 6, 2021 11.59am EDT

Dawn Skelton
Professor in Ageing and Health, Glasgow Caledonian University

Research shows that people’s ability to stand on one leg is an indicator of health and that getting
better at standing on one leg can add to fitness and potentially lifespan.

Being able to stand on one leg is linked to increased levels of physical activity and decreased risk of
falls and is associated with both quality and length of life. Around 37.3 million falls per year
worldwide are severe enough to require medical attention.

The inability to balance on one leg for 20 seconds or longer is linked in otherwise healthy people to an
increased risk of small blood vessel damage in the brain and reduced ability to understand ideas. You
are less likely to be able to stand on one leg without a wobble if you have a multitude of medical
conditions such as Parkinson’s disease, stroke or Alzheimer’s disease).

Pregnancy, menopause, the diagnosis of disease and retirement can also alter our strength and
balance and ability to stay upright, mostly because of the way these affect our ability and motivation
to engage in regular physical activity.
Sitting or reclining while awake is associated with lower muscle strength, risk of falls and physical
function, sometimes irrespective of the amount of moderate or vigorous intensity physical activity you
do. People who sit for prolonged periods are more likely to develop type 2 diabetes, higher waist
circumference and obesity.

Why standing on one leg matters

Data shows that the time people can stand on one leg changes for men and women at different ages –
and an inability to hit those target times for your age group can highlight health problems.

Read more:
Explainer: why does our balance get worse as we grow older?

The human body, when standing upright, is inherently unstable. We have a very small base of support
relative to our height and width. When in good health we rely on our central and peripheral nervous
system to integrate all the information coming in from our balance senses (eyes, inner ears and
feedback from muscles and joints). We then engage the right muscles (feet, ankle, leg and core
muscles, sometimes even the arm muscles) at the right time to make the necessary adjustments to our
posture to stay upright.

The more physically active we are, the more likely we are to have good balance and of course the wider
physical, psychological and social benefits of being regularly active are well known.

People doing an exercise class

More time standing on one leg or doing hopping exercises can improve your balance.
Rawpixel.com/Shutterstock

Other health problems can also affect our physical activity and are characterised by stooped posture
and worsened balance, such as depression and osteoporosis, leading to an increased risk of falls and
fracture. A stooped posture makes standing on one leg more difficult.

Should I start standing on one leg?

You can improve your ability to stand on one leg at any age, and improve your balance. It’s more
challenging if you have a neurological condition, but studies have shown improvements in balance
following exercise in many conditions and it is known to reduce falls in many others. In the UK
around one in three adults over 65 and half of people over 80 will have at least one fall a year. A
recent Public Health England report suggests that lack of activity in older adults because of the social
restrictions during the pandemic could lead to 110,000 more older people falling this year.

You can test your own balance here and have a try at some simple balance exercises. You are more
likely to have improved confidence in performing physical activity, and getting more active can help
lengthen and improve quality of life.
Standing on one leg can be made much more challenging if you close your eyes while doing it. This is
worth practising as it helps with our long- term balance and reduces the risk of falling over. With our
eyes open the body becomes lazy and relies on our vision to help keep us balanced. If you take out
vision, it gets the other senses working better. Most people can only do this for a short time compared
to when they keep their eyes open. Have something close by to hold on to. You will improve over time.

For the over 70s, there’s also evidence to show standing on each leg for one minute three times a day
can help improve hip bone mineral density. Stronger hip bone mineral density means if you do fall
you are less likely to fracture.

And on that note, do I stand on one leg? Yes, when I am cleaning my teeth morning and evening,
alternating from one leg to another at least twice. If I am feeling bold I try one round of the teeth with
my eyes closed. That, I might add, I am starting to find more difficult – I must practice more.