Module 2 case study
Case Study1: Medical Test Accuracy and Bayes’ Theorem
Problem Statement:
A new COVID-19 test has been introduced, and the manufacturer claims the following
probabilities:
The probability that a person has COVID-19 in the population is 5%.
If a person has COVID-19, the test correctly detects it (True Positive rate = 95%).
If a person does not have COVID-19, the test gives a False Positive (5%
probability).
A randomly selected individual tests positive for COVID-19. What is the probability that the
person actually has COVID-19? (Use Bayes' Theorem)
Tasks:
1. Compute the probability using Bayes' Theorem.
2. Interpret the result in terms of medical testing reliability.
3. If the false positive rate is reduced to 1%, how does the probability change?
Solution:
Bayes' Theorem
where:
P(A)= Probability of having COVID-19 = 5% (0.05)
P(B∣A)= Probability of testing positive given the person has COVID-19 (True
Positive Rate) = 95% (0.95)
P(Ac) = Probability of NOT having COVID-19 = 1 - 0.05 = 95% (0.95)
P(B∣Ac) = Probability of testing positive given the person does NOT have COVID-19
(False Positive Rate) = 5% (0.05)
Step 1: Compute the Probability
Thus, the probability that a person actually has COVID-19 given that they tested positive is
50% (0.5 or 50%).
Step 2: Interpretation of Medical Test Reliability
Even though the test has a high true positive rate (95%), the false positive rate
(5%) is relatively high.
As a result, when a person tests positive, there is only a 50% chance they actually
have COVID-19.
This means that false positives are common, so a second confirmatory test is
essential.
This demonstrates a common issue with medical tests: even highly accurate tests
can have low predictive power in populations with a low disease prevalence.
Step 3: Effect of Reducing the False Positive Rate
If the false positive rate is reduced to 1%, i.e., P(B∣Ac)=0.01, we recalculate:
Thus, if the false positive rate is reduced to 1%, the probability that a person actually has
COVID-19 given that they tested positive increases to 83.3%.
Case Study 2: Customer Behaviour and Conditional Probability
Problem Statement:
A shopping mall wants to analyse customer behaviour. Based on past data:
40% of customers visit the electronics section.
60% of customers visit the clothing section.
25% of customers visit both electronics and clothing sections.
If a customer is known to have visited the electronics section, what is the probability that
they also visited the clothing section? (Use Conditional Probability)
Tasks:
1. Construct a probability table for customer movements.
2. Use conditional probability to calculate the required probability.
3. Interpret the results and suggest how the mall can use this data for marketing.
Solution:
Step 1: Given Data
P(E) = Probability of visiting electronics section = 40% = 0.40
P(C) = Probability of visiting clothing section = 60% = 0.60
P(E ∩ C) = Probability of visiting both electronics and clothing = 25% = 0.25
We need to find the conditional probability that a customer visits the clothing section given
that they have already visited the electronics section. This is represented as:
P(C∣E)=P(E∩C)/P(E)
Step 2: Constructing the Probability Table
We create a joint probability table based on the given data:
Customer Behaviour Visited Electronics Did Not Visit Total
(E) Electronics (¬E)
Visited Clothing (C) 0.25 (Given) 0.60 - 0.25 = 0.35 0.60
(Given)
Did Not Visit Clothing 0.40 - 0.25 = 0.15 1 - (0.25 + 0.35 + 0.15) = 0.40
(¬C) 0.25 (Given)
Total 0.40 (Given) 0.60 1.00
Step 3: Compute Conditional Probability
Using the conditional probability formula:
P(C∣E) = P(E∩C)/P(E)
=0.25/0.40 = 0.625
Thus, if a customer visits the electronics section, there is a 62.5% probability that they
also visit the clothing section.
Step 4: Interpretation & Marketing Strategy
Interpretation of Results
62.5% of electronics customers also visit clothing → There is a strong relationship
between the two sections.
Only 37.5% of electronics customers do not visit clothing, meaning many
customers shop in both sections.
Case Study 3: Defective Products in a Manufacturing Process (Poisson Distribution)
Problem Statement:
A factory produces electronic chips. It has been observed that on average, 2 defective chips
occur per hour in a production line. The defects follow a Poisson distribution.
Tasks:
1. What is the probability that exactly 3 defective chips occur in an hour?
2. What is the probability that at most 1 defective chip occurs in an hour?
3. If the factory runs for 8 hours, what is the probability that the total number of
defective chips is greater than 20?
4. Interpret whether the production process is efficient based on the results.
Solution:
Poisson distribution formula:
P(X=k)=e−λλk/k!
where:
λ= average number of occurrences in a given time period
=2
k = specific number of occurrences,
e≈2.718
1. Probability of Exactly 3 Defective Chips in an Hour
Using the Poisson formula:
P(X=3) =e−223/3!
=e−2×8/6 ≈0.1804
So, the probability of exactly 3 defective chips in an hour is 0.1804 (18.04%).
2. Probability of At Most 1 Defective Chip in an Hour
“At most 1 defective chip” means P(X≤1) which is:
P(X=0)+P(X=1)
P(X=0) =e−220/0! = e−2=0.1353
P(X=1)=e−221/1! =0.2707
Summing them:
P(X≤1)=0.1353+0.2707=0.4060
So, the probability of at most 1 defective chip in an hour is 0.4060 (40.60%).
3. Probability That the Total Number of Defective Chips in 8 Hours is Greater
Than 20
For 8 hours, the new Poisson mean becomes:
λ′=8×2=16
We need to find P(X>20), which is:
1−P(X≤20)
Using a Poisson cumulative probability table or normal approximation:
P(X≤20)≈0.897
P(X>20)=1−0.897=0.103
So, the probability that more than 20 defective chips occur in 8 hours is 0.103
(10.3%).
4. Interpretation of Results
The probability of 3 defective chips in an hour is 18.04%, which is not too high.
The probability of at most 1 defective chip is 40.6%, meaning defects are relatively
rare.
Over 8 hours, there is only a 10.3% chance of exceeding 20 defects, indicating the
production process is reasonably efficient.
If maintaining low defects is critical, process improvement methods (such as quality
checks or machine calibration) should be considered to reduce defects further.
Case Study 4: Vehicle Speeds and Normal Distribution
Problem Statement:
The speeds of vehicles passing a highway checkpoint follow a Normal distribution with:
Mean speed = 70 km/h
Standard deviation = 10 km/h
A new law sets the speed limit at 80 km/h. The police want to find out:
1. What percentage of vehicles are exceeding the speed limit?
2. What percentage of vehicles are within the speed range of 60-80 km/h?
3. If a vehicle is randomly chosen, what is the probability that it is traveling at less than
50 km/h?
Tasks:
Use the Normal distribution function to calculate probabilities.
Interpret results in the context of traffic management and law enforcement.
Solution:
Z=X−μ /σ
where:
X = given speed value,
μ=70 km/h (mean speed),
σ=10 km/h (standard deviation).
We will then use Z-tables (Standard Normal Distribution tables) to find probabilities.
1. Probability of Vehicles Exceeding 80 km/h
We need to find:
P(X>80)
Step 1: Calculate the Z-score for 80 km/h:
Z=80−70/10=10/10=1
Step 2: Find the probability using the Z-table:
P(X<80) =P(Z<1) =0.8413
P(X > 80) = 1 - P(Z < 1) = 1 - 0.8413 = 0.1587
So, 15.87% of vehicles are exceeding the speed limit.
2. Probability of Vehicles Traveling Between 60 and 80 km/h
We need to find:
P(60≤X≤80)
Step 1: Calculate the Z-score for 60 km/h:
Z=60−70/10=−10/10=−1
Step 2: Use the Z-table:
P(X<60) =P(Z<−1) =0.1587
We already found:
P(X<80) =0.8413
Step 3: Compute the probability:
P(60≤X≤80) = P(X< 80) −P (X<60)
=0.8413−0.1587
=0.6826
So, 68.26% of vehicles are traveling within the speed range of 60-80 km/h.
Case Study 5: Call Centre Performance and Binomial Distribution
Problem Statement:
A call centre receives an average of 10 customer calls per hour. Each agent answers calls
independently, and the probability of successfully resolving a customer issue in a single call
is 70%.
Tasks:
1. What is the probability that exactly 7 out of 10 calls are successfully resolved?
2. What is the probability that at least 8 calls are successfully resolved?
3. If a new training program increases the success rate to 80%, how does the
probability change?
4. Interpret the results to assess the call centre’s performance and the impact of the
training program.
Solution:
Binomial Distribution formula:
where:
n = total number of trials (calls)
k = number of successful outcomes (resolved calls)
p = probability of success (resolving a call)
1. Probability that exactly 7 out of 10 calls are successfully resolved
Given:
n=10, k=7, p=0.7
= 0.2668
2. Probability that at least 8 calls are successfully resolved
We need to calculate:
P(X≥8 )=P(X=8)+P(X=9)+P(X=10) = 0.1937+0.1211+0.0282=0.343 or 34.3%
3. Effect of increasing the success rate to 80%
If p=0.8, we recalculate P(X=7) and P(X≥8).
=0.2013
P(X≥8) = P(x=8,9,10) =0.3019+ 0.2684+ 0.6777
Interpretation of Results
Initially, the probability of resolving at least 8 calls out of 10 was 34.3%.
After the training program (with p=0.8p = 0.8p=0.8), this probability increased to
67.77%.
This shows a significant improvement in call resolution, nearly doubling the success
rate for at least 8 resolved calls.
The training program clearly enhances the performance of call center agents.
This improvement can lead to higher customer satisfaction, reduced callback rates,
and better efficiency for the call center.
