SCNC1111

Scientific Method and Reasoning

Lecture 5
Mathematical Modelling

Dr. William M.Y. Cheung

1
 Previously on SCNC1111…
How to come up with
a non-ambiguous
theory?
Observation
What do you
(Generalization mean by “being
via Hypothesis non-ambiguous”?
induction ,
or Abduction)
Prediction
(Deduction)

Experimental
Test

A good theory Confirmation

must not be or Falsification
ambiguous!
 Also Previously on SCNC1111…
• Mathematics help us to be concrete
and precise!
– It would be difficult to be less
ambiguous when you tell me
 “My age is 18 year and 36 days ± 1 day.”
 “The balance of my bank account (S)
changes with time (t), following the
equation 𝑆 = 5000 sin 𝑡 + 5000”.
 Being Non-ambiguous via Being Mathematical
• So we should
1. express our hypothesis in terms of
equations and numerical parameters
e.g. 𝑆 = 5000 sin 𝑡 + 5000
2. make numerical predictions
e.g. “The balance of William’s bank account is
going to be $123.4  $0.1 on 29 February, 2020.”
3. compare with numerical
measurements from experiments
e.g. go and check the exact balance of
William’s bank account on 29 February, 2020
 A Potential Problem
• Not everything in our lives comes with numbers and
equations
e.g. marriage and divorce

Extracted from Death Note

 A Potential Problem
• Not everything in our lives comes with numbers and
equations
e.g. marriage and divorce
What factors lead
to divorce?

In what way do these factors

influence relationship?
Can we make concrete
predictions?

Let’s make a
Mathematical Model to
describe all these
phenomena!
 Models
• What comes to your mind when we mention “model”?

Fashion models?
Scale models?

Animal models for

disease studies?

• What properties do they share in common?

 Features of Models
• In all the above instances, the model is an
idealized representation or example for
thinking about a real-life phenomenon, such
as
– Wearing clothes;
– Manufacturing aeroplanes, robots;
– Town planning and building construction;
– Human diseases.
 Mathematical Modelling
• “Mathematical modelling is the description
of an experimentally delineated
phenomenon by means of mathematics,
with a view to capturing the salient aspects
of the phenomenon at hand.”
– Hugo van den Berg. Mathematical Models of
Biological Systems, Oxford, 2011. Page 3
• Not all aspects need to be captured!
– E.g. if you are only concerned about how the
population of cows changes with time,
– then it is OK to assume all cows are spherical!
Source: openlibrary.org Source: openlibrary.org
 How to Come Up with Mathematical Models
1. Stare hard at the problem:
– identify important aspects;
– make appropriate assumptions;
– decide on the key quantities our
model should describe;
– (equally importantly) forget about
non-essential factors.
 E.g. if you are wondering how water
flows in the Three Gorges project,
 you will not care about the water
molecules.
 How about Marriages?
• Seminal work by Professor John Gottman
of the University of Washington in the
1980s and 1990s
• His prediction of which couples would
divorce within a 4-year period is 94%
accurate.
• His team stared hard at the problem:
– They interviewed lots of couples;
– The interviews were videotaped;
– How a couple interacts with each other was
then analyzed.
 Important Aspects
• Important aspects
– Some previous studies considered gender
differences in communication styles cause
divorces.
– Gottman suggested how a couple
influences each other is more important.
– Whether one has a positive personality
also plays a role.
– Anger is not the most destructive emotion
in marriage:
 Happy couples also fight.
 Key Quantity and Assumptions
• Key quantity
– Happiness as a function of time, t!
 Happiness of husband, x(t)
 Happiness of wife, y(t)
 Negative value means you are unhappy!
• Assumptions
– Everyone tends to his/her own “intrinsic happiness value”
when one is single.
 Call it 𝑥0 or 𝑦0 .
 Its value is larger for a more positive person.
– When in a relationship, one’s happiness is increased or
decreased based on the partner’s happiness.
 Different person reacts differently.
 Now What?
2. Translate your problem into a
mathematical one:
– describe how the key quantities
affect one another
 Draw a flow chart!
– using mathematical tools that can
help you to draw conclusions.
 Flip through your mathematics
textbook as there can be more than
one approaches!
 How to Handle Dice-throwing
• We are used to treating the results of
throwing a die as probabilistic:
– a chance of 1/6 to get 1, 2, 3, 4, 5, or 6
• However mechanics informs us that the
result can be determined with certainty
– if we are given all the necessary information,
such as
 the force we apply;
 the air current nearby;
 how the die bounces.
 How to Handle Dice-throwing
• Too complicated!
– And we won’t get all the necessary
information anyway.
– Let’s revert back to Probability.
• So the approach adopted depends on
what we would like to achieve!
‒ We need not be entirely correct,
‒ since we ignore irrelevant aspects.
 Gottman’s Model
• We would like to study how a couple’s happiness changes
with time.
‒ Calculus and Differential Equations are great tools to handle
these situations.
‒ Also why these topics are included in our course
‒ But we will not care about the mathematical details for the
moment. Describe how one
approaches one’s
𝑑𝑥 “intrinsic happiness
= 𝑟ℎ 𝑥0 − 𝑥 + 𝐼ℎ 𝑦 value” when single.
Change in happiness
𝑑𝑡
per unit time
𝑑𝑦 Describe how one’s
= 𝑟𝑤 𝑦0 − 𝑦 + 𝐼𝑤 (𝑥) happiness is influenced
𝑑𝑡 by that of one’s partner
 Gottman’s Influence Functions
𝑎ℎ 𝑦 if 𝑦 > 0
𝐼ℎ 𝑦 =
𝑏ℎ 𝑦 if 𝑦 < 0
• Different person reacts differently to partner’s positive and
negative emotion.
• E.g. Conflict avoider
10𝑦 if 𝑦 > 0
𝐼ℎ 𝑦 =
0.1𝑦 if 𝑦 < 0

Do not react to partner’s Enjoy happiness

negative emotion with partner
 Other Influence Styles
• E.g. Reinforcer
10𝑦 if 𝑦 > 0
𝐼ℎ 𝑦 =
9𝑦 if 𝑦 < 0
Enjoy happiness
together
Induce unhappiness
together

• E.g. Pessimist

0.1𝑦 if 𝑦 > 0
𝐼ℎ 𝑦 = Do not react to
partner’s positive
10𝑦 if 𝑦 < 0 emotion
Induce unhappiness
together
 Parameter Estimation
3. Estimate numerical values of the
parameters in our model
− E.g. 𝑎ℎ , 𝑏ℎ , 𝑟ℎ in Gottman’s model
by
− guesstimating;
− making reference to known results;
− or direct measurements.
 Direct measurements involve handling large
amount of data.
 This is why we also need to know some Statistics.
− Sometimes this stage is postponed,
depending on what we would like to
achieve.
 Regression
y

C
B

Which line fits the data better?

Imagine the slope of this line gives
us the value of our parameter…
 Parameter Estimation
3. Estimate numerical values of the
parameters in our model
− E.g. 𝑎ℎ , 𝑏ℎ , 𝑟ℎ in Gottman’s model
by
− guesstimating;
− making reference to known results;
− or direct measurements.
 Direct measurements involve handling large
amount of data.
 This is why we also need to know some Statistics.
− Sometimes this stage is postponed,
depending on what we would like to
achieve.
 What did Gottman do?
• During the interview of a couple
− The couple discussed certain topic.
− Each of their comments was then rated as
positive or negative.
− One’s positivity/negativity was quantified by
the number of positive/negative comments
one made.
− One’s positivity/negativity was then
compared with that of the other using
Statistics.
 Parameters for “Singular” Behaviour
• How about parameters that describe how one
behaves when single?
‒ i.e. 𝑥0 , 𝑦0 , 𝑟𝑤 and 𝑟ℎ
• Observe one’s behaviour when the partner is
being neutral
‒ i.e. no. of positive comments made by partner 
no. of negative comments made by partner
‒ 𝐼ℎ or 𝐼𝑤 ≈ 0
‒ about 15% of Gottman’s data
 Looking for Solution
4. “Solve” the mathematical problem
i.e. what the mathematical model
can tell us about the phenomenon
under study via
‒ finding analytic solutions
‒ performing numerical simulations
 Can make use of the online computer
algebra software, Wolfram Alpha
‒ In Gottman’s model, we would like to
find the long-term happiness after
marriage
 Call it 𝑥∗ and 𝑦∗
Wolfram Alpha
 Looking for Solution
4. “Solve” the mathematical problem
i.e. what the mathematical model
can tell us about the phenomenon
under study via
‒ finding analytic solutions
‒ performing numerical simulations
 Can make use of the online computer
algebra software, Wolfram Alpha
‒ In Gottman’s model, we would like to
find the long-term happiness after
marriage
 Call it 𝑥∗ and 𝑦∗
 Solution Interpretation
5. Interpret the conclusions we
obtain from the model
− i.e. what does our mathematical
solution tell us about the real-
life phenomena?
− In Gottman’s model,
 if 𝑥∗ < 𝑥0 , then the husband
becomes less happy by getting into
marriage…
 some day he may wake up, and
suggests divorce to his wife.
 The same for the wife…
 Model Verification
6. Verify the model
• In this stage the strengths and weaknesses of the
model should be examined.

• “All models are wrong, but

some are useful”
– George E.P. Box (1919 – 2013)
George E.P. Box
Source: Wikipedia

• We should not require a model to be “correct”;

• but it should be valid for our study of the phenomena.
 The Modelling Process
Decide important Translate these
aspects of a important aspects into Obtain numerical
phenomenon mathematical problem values of the
necessary
parameters
Solutions of model disagree with phenomenon: Verify the model:
use this to modify the model how does it
compare with
reality?
Solutions of model agree Solve mathematical
with phenomenon problem

Use the model to make Interpret the solution

further predictions and
pose relevant questions Perform an experiment

Science Mathematics

1. An interplay between Science and Mathematics!

2. Use the model to suggest new insights into the phenomenon!