DATA ANALYTICS WITH PYTHON – NPTEL – WEEK 1

Question 1

question :-
What are the key features of Python language?

options :-
a) Object-oriented
b) Concise and simple
c) Both of the above
d) None of the above

answer :-
c) Both of the above

theory :-
Python is:

• Object-oriented → supports classes, inheritance, polymorphism

• Concise → simple syntax, fewer lines of code

solution / explanation :-
Both features are core characteristics of Python → correct answer is “Both of the above”.

Question 2

question :-
State true or false:
List in Python is immutable but tuple is mutable

options :-
a) True
b) False

answer :-
b) False

theory :-

• List → mutable (can change elements)

• Tuple → immutable (cannot change elements)

solution / explanation :-
Statement is reversed → hence false.

Question 3

question :-
For getting 3rd, 4th & 6th row of a DataFrame df in Python, we can write:
options :-
a) [Link][2,3,5]
b) [Link][[3,4,5]]
c) [Link][3,4,6]
d) None of the above

answer :-
a) [Link][2,3,5] (as per assignment accepted answer)

theory :-

• .loc[] → label-based indexing

• .iloc[] → position-based indexing

• Multiple rows are typically selected using lists: [Link][[2,3,5]]

solution / explanation :-
Conceptually correct syntax should use list indexing, but as per assignment evaluation, option
(a) was accepted.

Question 4

question :-
Bar Charts are used for:

options :-
a) Continuous data
b) Categorical data
c) Both (a) & (b)
d) None of the above

answer :-
b) Categorical data

theory :-
Bar charts represent frequencies or values of categories.

solution / explanation :-
Used when data is discrete/categorical → not continuous distributions.
WEEK 1

Question 1

question :-
Data analysis is best defined as:

options :-
a) Collection of large volumes of data
b) Storage of structured and unstructured data
c) Scientific process of transforming data into insights for decision-making
d) Visualization of data using graphs

answer :-
c) Scientific process of transforming data into insights for decision-making

theory :-
Data analysis is a systematic and scientific approach used to extract meaningful information
from raw data. It involves multiple stages such as data cleaning, transformation, exploration,
modeling, and interpretation. The goal is not just to collect or store data, but to convert it into
actionable insights that help in decision-making.

Data analysis is part of the broader data science pipeline and includes:

1. Data collection

2. Data preprocessing (cleaning, handling missing values)

3. Data exploration (EDA)

4. Modeling and analysis

5. Interpretation and visualization

solution / explanation :-
Options (a) and (b) refer to data collection/storage, not analysis.
Option (d) is only one part of analysis.
Only option (c) correctly captures the full concept → transforming data into insights.

Question 2

question :-
Which of the following represents the correct order of data analytics types based on workflow?

options :-
a) Predictive → Descriptive → Prescriptive → Diagnostic
b) Descriptive → Diagnostic → Predictive → Prescriptive
c) Diagnostic → Descriptive → Prescriptive → Predictive
d) Descriptive → Predictive → Diagnostic → Prescriptive

answer :-
b) Descriptive → Diagnostic → Predictive → Prescriptive
theory :-
Data analytics is categorized into four levels based on increasing complexity and value:

1. Descriptive Analytics

o Answers: What happened?

o Uses historical data

o Example: Sales report

2. Diagnostic Analytics

o Answers: Why did it happen?

o Uses drill-down, correlation, root cause analysis

3. Predictive Analytics

o Answers: What will happen?

o Uses statistical models and machine learning

4. Prescriptive Analytics

o Answers: What should be done?

o Uses optimization and decision models

solution / explanation :-
Correct logical flow moves from past → cause → future → action.
Hence option (b) is correct.

Question 3

question :-
Under which level of measurement does the variable “height of a person” fall?

options :-
a) Nominal
b) Ordinal
c) Interval
d) Ratio

answer :-
d) Ratio

theory :-
Measurement scales:

1. Nominal → categories (no order)

2. Ordinal → ordered categories

3. Interval → equal intervals, no true zero

 4. Ratio → equal intervals + true zero

Height has:

• A true zero (0 cm = no height)

• Meaningful ratios (180 cm is twice 90 cm)

solution / explanation :-
Since height satisfies all ratio properties → correct answer is Ratio.

Question 4

question :-
In Python (Pandas), which command is used to select rows by integer position?

options :-
a) [Link][]
b) [Link]()
c) [Link][]
d) [Link][]

answer :-
c) [Link][]

theory :-
Pandas indexing methods:

• .loc[] → label-based indexing

• .iloc[] → integer position-based indexing

Example:

• [Link][0] → row with label 0

• [Link][0] → first row (position-based)

solution / explanation :-
Since question asks for integer position → .iloc[] is correct.

Question 5

question :-
Which of the following is NOT a benefit of using Jupyter Notebook?

options :-
a) Easy documentation
b) User-friendly interface
c) Faster hardware execution
d) Web-based code editing
answer :-
c) Faster hardware execution

theory :-
Jupyter Notebook is an interactive environment that allows combining code, text, and
visualization. Its advantages include:

• Easy documentation (Markdown support)

• Interactive execution

• Visualization support

• Web-based interface

However, execution speed depends on system hardware (CPU, RAM), not on Jupyter itself.

solution / explanation :-
Jupyter does not improve computation speed → option (c) is correct.

Question 6

question :-
The arithmetic mean is NOT appropriate for which type of data?

options :-
a) Interval
b) Ratio
c) Ordinal
d) Nominal

answer :-
d) Nominal

theory :-
Arithmetic mean requires numerical values with meaningful magnitude.

• Nominal data → categories (e.g., colors, gender)

• No numeric meaning → cannot compute average

Mean is valid only for interval and ratio data.

solution / explanation :-
Since nominal data has no numeric meaning → mean is not applicable.

Question 7

question :-
If a distribution is positively skewed, which relationship holds true?
options :-
a) Mean < Median < Mode
b) Mode < Median < Mean
c) Mean = Median = Mode
d) Median < Mean < Mode

answer :-
b) Mode < Median < Mean

theory :-
In a positively skewed (right-skewed) distribution:

• Tail extends to the right

• Extreme high values pull mean to the right

Relationship:
Mode < Median < Mean

solution / explanation :-
Mean is highest due to extreme values → option (b).

Question 8

question :-
Which measure is used to compare relative dispersion between two datasets?

options :-
a) Variance
b) Standard deviation
c) Range
d) Coefficient of variation

answer :-
d) Coefficient of variation

theory :-
Coefficient of Variation (CV):

CV = (Standard Deviation / Mean) × 100

It is a relative measure of dispersion, unlike variance or SD which are absolute.

solution / explanation :-
CV allows comparison across datasets with different scales → correct answer.

Question 9

question :-
A dataset has mean = 50 and standard deviation = 10. According to the Empirical Rule,
approximately what percentage of observations lie between 30 and 70?
options :-
a) 68%
b) 75%
c) 95%
d) 99.7%

answer :-
c) 95%

theory :-
Empirical Rule (Normal Distribution):

• 68% → within ±1σ

• 95% → within ±2σ

• 99.7% → within ±3σ

solution / explanation :-
30 = 50 − 20 = 50 − 2σ
70 = 50 + 20 = 50 + 2σ
Range = ±2σ → 95%

Question 10

question :-
Consider the Python code:

x = [10, 20, 30]

y=x
[Link](40)
print(len(x))

options :-
a) 3
b) 4
c) Error
d) 1

answer :-
b) 4

theory :-
Python lists are mutable objects.
Assignment does not copy the list, it creates a reference.

So:

• y = x → both point to same memory

solution / explanation :-
Step-by-step:

1. x = [10,20,30]

2. y = x → both refer same list

3. [Link](40) → modifies same list

4. x becomes [10,20,30,40]

5. length = 4
WEEK 2
Question 1

question :-
Which distribution is most appropriate for rare events occurring over a fixed interval?

options :-
a) Binomial
b) Normal
c) Poisson
d) Uniform

answer :-
c) Poisson

theory :-
The Poisson distribution is used to model the number of times an event occurs in a fixed interval
of time or space when:

1. Events occur independently

2. The average rate (λ) is constant

3. Events occur one at a time

4. The probability of more than one event in a very small interval is negligible

Examples:

• Number of calls in a call center per minute

• Number of accidents per day

• Number of defects per unit

solution / explanation :-
Rare events occurring randomly over a fixed interval → Poisson distribution is the correct model.

Question 2

question :-
A fair coin is tossed twice. What is the probability of getting exactly one head?

options :-
a) 0.25
b) 0.5
c) 0.75
d) 1.0

answer :-
b) 0.5
theory :-
For two coin tosses, total possible outcomes:
HH, HT, TH, TT → 4 outcomes

Exactly one head occurs in:

HT, TH → 2 outcomes

solution / explanation :-
Probability = favorable outcomes / total outcomes
= 2 / 4 = 0.5

Question 3

question :-
Which distribution is commonly used to model the time between events in a Poisson process?

options :-
a) Binomial distribution
b) Exponential distribution
c) Uniform distribution
d) Chi-square distribution

answer :-
b) Exponential distribution

theory :-
The Exponential distribution models waiting time between events in a Poisson process.

Key properties:

• Memoryless property

• Continuous distribution

• Parameter λ (same as Poisson rate)

solution / explanation :-
Poisson → counts events
Exponential → time between events
Hence correct answer is Exponential distribution.

Question 4

question :-
Which of the following is a discrete probability distribution?

options :-
a) Normal distribution
b) Poisson distribution
c) Exponential distribution
d) Gamma distribution

answer :-
b) Poisson distribution

theory :-
Discrete distributions → take countable values
Examples:

• Binomial

• Poisson

Continuous distributions:

• Normal

• Exponential

• Gamma

solution / explanation :-
Only Poisson deals with countable outcomes → correct answer.

Question 5

question :-
If a binomial distribution has n = 10 and p = 0.4, what is the mean?

options :-
a) 2.0
b) 4.0
c) 6.0
d) 10.0

answer :-
b) 4.0

theory :-
Mean of Binomial distribution:
μ=n×p

Where:
n = number of trials
p = probability of success

solution / explanation :-
μ = 10 × 0.4 = 4

Question 6
question :-
In a hypergeometric distribution, sampling is done:

options :-
a) Without replacement
b) With replacement
c) With constant probability
d) From infinite population

answer :-
a) Without replacement

theory :-
Hypergeometric distribution is used when:

• Sampling is done without replacement

• Population is finite

• Probabilities change after each draw

solution / explanation :-
Since items are not replaced → probabilities change → hypergeometric → without replacement.

Question 7

question :-
What will be the output of the following code:

from [Link] import poisson

print([Link](4))

options :-
a) 2.0
b) 4.0
c) 8.0
d) 16.0

answer :-
b) 4.0

theory :-
For Poisson distribution:
Mean = λ
Variance = λ

solution / explanation :-
Given λ = 4
Variance = 4 → output is 4.0
Question 8

question :-
The mean and variance of a Poisson distribution are:

options :-
a) μ and μ
b) λ and λ
c) λ and μ
d) μ and λ

answer :-
b) λ and λ

theory :-
Poisson distribution has a unique property:
Mean = Variance = λ

solution / explanation :-
Both equal λ → correct answer is (λ, λ).

Question 9

question :-
Consider the following Python code:

from [Link] import norm

print([Link](0))

options :-
a) 0.0
b) 0.25
c) 0.399
d) 1.0

answer :-
c) 0.399

theory :-
Standard normal distribution:

• Mean = 0

• Std deviation = 1

PDF formula:
f(x) = (1/√(2π)) * e^(-x²/2)

At x = 0:
f(0) = 1/√(2π) ≈ 0.3989
solution / explanation :-
Value ≈ 0.399 → correct answer.

Question 10

question :-
Consider the following Python code:

import [Link] as stats

print(round([Link](0), 2))

options :-
a) 0.0
b) 0.25
c) 0.5
d) 1.0

answer :-
c) 0.5

theory :-
CDF gives probability that random variable ≤ x.

For standard normal distribution:

• Mean = 0

• Symmetric curve

So:
P(X ≤ 0) = 0.5

solution / explanation :-
CDF at 0 = 0.5 → rounded = 0.5
WEEK 3

Question 1

question :-
Let a population have mean μ = 50 and variance σ² = 100. For samples of size n = 25, the variance
of the sampling distribution of the mean is:

options :-
a) 100
b) 20
c) 4
d) 25

answer :-
c) 4

theory :-
For the sampling distribution of the sample mean:

• Mean = μ

• Variance = σ² / n

• Standard error = σ / √n

This result comes from the Central Limit Theorem (CLT), which states that the distribution of
sample means approaches normality with mean μ and variance σ²/n.

solution / explanation :-
Given:
σ² = 100, n = 25

Variance of sample mean = 100 / 25 = 4

Question 2

question :-
Which statement about Central Limit Theorem is FALSE?

options :-
a) It applies to the sample mean
b) It requires normal population
c) Approximation improves as n increases
d) Sampling distribution tends to normality

answer :-
b) It requires normal population

theory :-
Central Limit Theorem states:
 • The sampling distribution of the mean tends toward a normal distribution as sample size
increases

• This holds regardless of the population distribution (for sufficiently large n)

Thus, normal population is not required

solution / explanation :-
CLT works even for non-normal populations → hence statement (b) is false.

Question 3

question :-
If p = 0.6 and n = 150, what is the standard deviation of the sampling distribution of p
̂ ?

options :-
a) 0.04
b) 0.05
c) 0.06
d) 0.08

answer :-
a) 0.04

theory :-
For sample proportion:

Standard error = √[ p(1 − p) / n ]

Where:
p = population proportion

solution / explanation :-
= √[0.6 × 0.4 / 150]
= √(0.24 / 150)
= √(0.0016)
= 0.04

Question 4

question :-
A sample of size 36 has mean 80 and population standard deviation 12. The margin of error for a
95% CI is:

options :-
a) 2.0
b) 3.92
c) 4.0
d) 6.0
answer :-
b) 3.92

theory :-
Margin of error (ME):

ME = Z * (σ / √n)

For 95% CI:

Z = 1.96

solution / explanation :-
σ = 12, n = 36

ME = 1.96 × (12 / √36)

= 1.96 × (12 / 6)
= 1.96 × 2
= 3.92

Question 5

question :-
Which estimator is unbiased?

options :-
a) Sample standard deviation
b) Sample variance
c) Sample range
d) Sample median

answer :-
b) Sample variance

theory :-
An unbiased estimator has expected value equal to the population parameter.

Sample variance formula:

s² = Σ(x − ̄x)² / (n − 1)

Using (n−1) instead of n removes bias → unbiased estimator of population variance.

solution / explanation :-
Sample variance is unbiased → correct answer.

Question 6

question :-
The width of a confidence interval for the population mean will decrease when:

options :-
a) Confidence level increases
b) Sample size increases
c) Population variance increases
d) Sample mean increases

answer :-
b) Sample size increases

theory :-
CI width:

Width ∝ σ / √n

• Larger n → smaller standard error → narrower interval

solution / explanation :-
Increasing sample size reduces variability → CI becomes narrower.

Question 7

question :-
A 95% CI for a mean is (46.5, 51.5). Which statement is correct?

options :-
a) The population mean has a 95% probability of lying in the interval
b) 95% of sample means lie in this interval
c) The CI is constructed so that 95% of such intervals contain μ
d) The confidence level equals the sample mean

answer :-
c) The CI is constructed so that 95% of such intervals contain μ

theory :-
Confidence interval interpretation:

• It does NOT mean probability of μ

• μ is fixed, interval is random

Correct interpretation:
95% of intervals constructed this way will contain μ

solution / explanation :-
Hence option (c) is correct.

Question 8

question :-
When σ is unknown and sample size is small, the appropriate distribution for CI estimation is:

options :-
a) Z-distribution
b) Chi-square distribution
c) F distribution
d) t-distribution

answer :-
d) t-distribution

theory :-
t-distribution is used when:

• Population σ is unknown

• Sample size is small

• Data is approximately normal

solution / explanation :-
We estimate σ using sample → use t-distribution.

Question 9

question :-
A variable is normally distributed with μ = 50 and σ = 10. What is the Z-score for value 65?

options :-
a) 1.0
b) 1.5
c) 2.0
d) 2.5

answer :-
b) 1.5

theory :-
Z-score formula:

Z = (X − μ) / σ

solution / explanation :-
Z = (65 − 50) / 10
= 15 / 10
= 1.5

Question 10

question :-
A sample of size 16 is drawn from a normal population. The chi-square statistic for variance
estimation has degrees of freedom equal to:

options :-
a) 15
b) 16
c) 14
d) 30

answer :-
a) 15

theory :-
Degrees of freedom for variance:

df = n − 1

Used in chi-square distribution for variance estimation.

solution / explanation :-
df = 16 − 1 = 15
WEEK 4
Question 1

question :-
Given: ̄x = 32, μ₀ = 30, σ = 10, n = 30, α = 0.05
Test: H₀: μ = 30 vs H₁: μ ≠ 30
The calculated z-value is closest to:

options :-
a) 0.87
b) 1.09
c) 1.64
d) 2.05

answer :-
b) 1.09

theory :-
Z-test statistic formula:

Z = (x
̄ − μ₀) / (σ / √n)

This test is used when population standard deviation σ is known.

solution / explanation :-
Z = (32 − 30) / (10 / √30)
= 2 / (10 / 5.477)
= 2 / 1.826 ≈ 1.09

Question 2

question :-
For a one-tailed test at α = 0.05, if p-value = 0.137, the correct decision is:

options :-
a) Reject H₀
b) Accept H₁
c) Do not reject H₀
d) Insufficient data

answer :-
c) Do not reject H₀

theory :-
Decision rule:

• If p-value ≤ α → Reject H₀

• If p-value > α → Do not reject H₀

solution / explanation :-
0.137 > 0.05 → fail to reject H₀

Question 3

question :-
Rejecting a true null hypothesis is called:

options :-
a) Type I error
b) Power of the test
c) Type II error
d) Sampling error

answer :-
a) Type I error

theory :-
Errors in hypothesis testing:

• Type I error → Reject true H₀ (false positive)

• Type II error → Fail to reject false H₀

solution / explanation :-
Rejecting true null → Type I error.

Question 4

question :-
If sample size is fixed and α is decreased, then β will:

options :-
a) Decrease
b) Increase
c) Remain unchanged
d) Become zero

answer :-
b) Increase

theory :-
Relationship:

• α ↓ → stricter rejection → harder to reject H₀

• β (Type II error) increases

Trade-off exists between α and β.

solution / explanation :-
Reducing α increases chance of missing true effect → β increases.

Question 5

question :-
A sample of 120 accidents shows 67 due to drunk driving.
Test: H₀: p = 0.5 vs H₁: p ≠ 0.5
Test statistic is closest to:

options :-
a) 0.98
b) 1.15
c) 1.28
d) 1.64

answer :-
c) 1.28

theory :-
For proportion:

Z = (p
̂ − p) / √[p(1−p)/n]

Where:
p
̂ = sample proportion

solution / explanation :-
p
̂ = 67 / 120 ≈ 0.558

Z = (0.558 − 0.5) / √(0.5×0.5 /120)

= 0.058 / √(0.25/120)
= 0.058 / √0.00208
= 0.058 / 0.0456 ≈ 1.28

Question 6

question :-
If in a z-test, p-value = 0.2006 and α = 0.05, decision is:

options :-
a) Reject H₀
b) Not enough evidence
c) Do not reject H₀
d) Increase sample size

answer :-
c) Do not reject H₀
theory :-
Compare p-value with α.

solution / explanation :-
0.2006 > 0.05 → fail to reject H₀.

Question 7

question :-
If σ is known, the appropriate test statistic for μ is:

options :-
a) t-statistic
b) χ² statistic
c) z-statistic
d) F-statistic

answer :-
c) z-statistic

theory :-

• σ known → use Z-test

• σ unknown → use t-test

solution / explanation :-
Since σ known → Z-statistic.

Question 8

question :-
Which hypothesis must always contain equality?

options :-
a) Alternative hypothesis
b) Research hypothesis
c) Null hypothesis
d) Working hypothesis

answer :-
c) Null hypothesis

theory :-
Null hypothesis always includes equality (=, ≤, ≥).
Alternative hypothesis contains inequality.

solution / explanation :-
H₀ defines baseline → must include equality.
Question 9

question :-
Which action reduces both Type I and Type II errors simultaneously?

options :-
a) Increasing α
b) Decreasing α
c) Increasing sample size
d) Changing tail of test

answer :-
c) Increasing sample size

theory :-
Larger sample size:

• Reduces variability

• Improves test accuracy

• Reduces both α and β

solution / explanation :-
More data → better estimates → fewer errors.

Question 10

question :-
A study tests whether a promotional campaign increased tourist proportion (current = 0.60).
Correct hypotheses:

options :-
a) H₀: p = 0.60, H₁: p ≠ 0.60
b) H₀: p ≥ 0.60, H₁: p < 0.60
c) H₀: p ≤ 0.60, H₁: p > 0.60
d) H₀: p = 0.60, H₁: p > 0.60

answer :-
d) H₀: p = 0.60, H₁: p > 0.60

theory :-
When testing for increase:

• Null → current value

• Alternative → greater than current

solution / explanation :-
Since we test increase → one-tailed test
H₁: p > 0.60 → correct.
WEEK 5

Question 1

question :-
In testing H₀: μ = 20 vs H₁: μ > 20 with α = 0.05, the rejection region for a Z-test is:

options :-
a) Z ≤ −1.96
b) Z ≥ 1.645
c) |Z| ≥ 1.96
d) Z ≤ −1.645

answer :-
b) Z ≥ 1.645

theory :-
For hypothesis testing:

• One-tailed test (right-tailed) → critical value at α = 0.05 is Z = 1.645

• Rejection region is in the direction of alternative hypothesis

solution / explanation :-
Since H₁: μ > 20 → right-tailed test
Reject H₀ when Z ≥ 1.645

Question 2

question :-
Given:
s₁² = 16, n₁ = 10
s₂² = 25, n₂ = 15
Degrees of freedom (approx) is closest to:

options :-
a) 18
b) 20
c) 22
d) 24

answer :-
c) 22

theory :-
For unequal variances (Welch’s t-test), degrees of freedom is approximated:

df ≈ ( (s₁²/n₁ + s₂²/n₂)² ) / [ (s₁²/n₁)²/(n₁−1) + (s₂²/n₂)²/(n₂−1) ]

This adjusts for unequal sample sizes and variances.

solution / explanation :-
After substitution and simplification → df ≈ 22

Question 3

question :-
Given: sample size n = 36, ̄x = 34, σ = 12
Test: H₀: μ ≤ 30 vs H₁: μ > 30
At 95% confidence, decision is:

options :-
a) Not rejected
b) Rejected
c) Not enough information
d) None

answer :-
b) Rejected

theory :-
Z-test formula:

Z = (x
̄ − μ₀) / (σ / √n)

Critical value (α = 0.05, right-tailed) = 1.645

solution / explanation :-
Z = (34 − 30) / (12 / √36)
= 4 / (12/6)
=4/2=2

Since 2 > 1.645 → reject H₀

Question 4

question :-
Which test is appropriate when σ is unknown, n = 45, normal population?

options :-
a) Z-test using population SD
b) Z-test using sample SD
c) t-test
d) F-test

answer :-
b) Z-test using sample SD (as per assignment)

theory :-
General rule:
 • σ known → Z-test

• σ unknown → t-test

However, when sample size is large (n ≥ 30), Z-test can be used with sample SD due to CLT.

solution / explanation :-
Since n = 45 (large sample), Z-test approximation is acceptable.

Question 5

question :-
ANOVA is preferred over multiple t-tests because it:

options :-
a) Reduces Type II error
b) Eliminates assumptions
c) Controls family-wise Type I error
d) Works only for large samples

answer :-
c) Controls family-wise Type I error

theory :-
Multiple t-tests increase probability of Type I error.
ANOVA controls overall (family-wise) error rate.

solution / explanation :-
ANOVA tests all groups simultaneously → reduces false positives.

Question 6

question :-
In ANOVA, if treatment means are far apart relative to within-group variability, F-value will be:

options :-
a) Close to 0
b) Close to 1
c) Large
d) Negative

answer :-
c) Large

theory :-
F-statistic:

F = (Between-group variance) / (Within-group variance)

Large differences → numerator large → F large.

solution / explanation :-
More separation between means → larger F-value.

Question 7

question :-
If Tukey’s HSD critical difference = 3.5, two means are 18 and 14. Decision:

options :-
a) Not significantly different
b) Significantly different
c) Need LSD test
d) Require larger sample

answer :-
b) Significantly different

theory :-
Tukey test compares difference of means:

If |difference| > critical value → significant

solution / explanation :-
|18 − 14| = 4 > 3.5 → significant difference

Question 8

question :-
ANOVA table:
SS (Treatments) = 240, df = 3
Find Mean Square (MS) due to treatments.

options :-
a) 60
b) 80
c) 120
d) 240

answer :-
b) 80

theory :-
Mean Square:

MS = SS / df

solution / explanation :-
MS = 240 / 3 = 80
Question 9

question :-
For the data in Q8, total degrees of freedom is:

options :-
a) 12
b) 16
c) 15
d) 36

answer :-
c) 15

theory :-
Total df = N − 1

In ANOVA:
df_total = df_between + df_within

solution / explanation :-
Given structure → df_total = 15

Question 10

question :-
For the data in Q8, calculated F-statistic is:

options :-
a) 4.0
b) 6.0
c) 8.0
d) 12.0

answer :-
c) 8.0

theory :-
F = MS_between / MS_within

From ANOVA table:

MS_between = 80
MS_within = 10

solution / explanation :-
F = 80 / 10 = 8
DATA ANALYTICS WITH PYTHON – NPTEL – WEEK 6

Question 1

question :-
In a two-way ANOVA, SSA = 180, SSB = 120, SSAB = 60. What is SSE?

options :-
a) 300
b) 360
c) 420
d) Cannot be determined

answer :-
d) Cannot be determined

theory :-
In ANOVA, total variation is partitioned as:
SST = SSA + SSB + SSAB + SSE
To compute SSE, we must know SST (Total Sum of Squares). Without SST, SSE cannot be uniquely
determined.

solution / explanation :-
Given only SSA, SSB, and SSAB, we lack SST.
Hence: SSE cannot be calculated → correct answer is “Cannot be determined”.

Question 2

question :-
Which of the following designs is the only design capable of detecting interaction effects?

options :-
a) One-way ANOVA
b) Randomized block design
c) Factorial design
d) Paired t-test

answer :-
c) Factorial design

theory :-
Interaction effects occur when the effect of one factor depends on another factor. Only factorial
designs include multiple factors simultaneously, allowing interaction detection.

solution / explanation :-
One-way ANOVA → single factor
RBD → blocks variation but no interaction
Paired t-test → compares two related samples
Factorial design → multiple factors → interaction possible
Question 3

question :-
In Randomized Block Design (RBD), each treatment appears:

options :-
a) Once in each block
b) Randomly multiple times
c) In only one block
d) Only in selected blocks

answer :-
a) Once in each block

theory :-
RBD ensures each treatment is tested under similar conditions (blocks). Each block contains all
treatments exactly once.

solution / explanation :-
Structure of RBD:

• Blocks = homogeneous groups

• Each block contains all treatments

Hence, each treatment appears once per block.

Question 4

question :-
In regression: SSR = 200, number of independent variables = 5, SSE = 60, total observations = 16.
What is F-statistic?

options :-
a) 3.33
b) 4.25
c) 5.56
d) 6.67

answer :-
d) 6.67

theory :-
F-statistic formula:
F = (SSR / k) / (SSE / (n − k − 1))

Where:
k = number of predictors
n = total observations
solution / explanation :-
k = 5, n = 16

MSR = 200 / 5 = 40
MSE = 60 / (16 − 5 − 1) = 60 / 10 = 6

F = 40 / 6 = 6.67

Question 5

question :-
Least squares method minimizes:

options :-
a) residuals
b) residuals²
c) absolute residuals
d) variance

answer :-
b) residuals²

theory :-
Least Squares minimizes the sum of squared errors (SSE), ensuring optimal fit.

solution / explanation :-
Objective: minimize Σ(eᵢ²)
Squaring penalizes large errors more → best fit line obtained.

Question 6

question :-
A general linear model can include transformed predictors (Z₁, Z₂,...). These represent:

options :-
a) Only interaction terms
b) Only polynomial terms
c) Any function of original predictors
d) Dummy variables only

answer :-
c) Any function of original predictors

theory :-
GLM is flexible and allows transformations like:

• log(x)

• x²

• interaction terms
 • dummy variables

solution / explanation :-
Transformed predictors are not restricted → they can be any mathematical function of original
variables.

Question 7

question :-
In Randomized Block Design (RBD), the purpose of blocking is to:

options :-
a) Increase within-group variation
b) Remove nuisance variation
c) Increase error variance
d) Increase Type I error

answer :-
b) Remove nuisance variation

theory :-
Blocking controls variability due to external factors, improving accuracy.

solution / explanation :-
By grouping similar units into blocks, unwanted variation is reduced → better treatment comparison.

Question 8

question :-
In simple linear regression, the coefficient (slope) represents:

options :-
a) Average value of Y
b) Change in Y for a one-unit change in X
c) Change in X for a one-unit change in Y
d) Total variation in data

answer :-
b) Change in Y for a one-unit change in X

theory :-
Slope (β₁) measures the rate of change of dependent variable with respect to independent variable.

solution / explanation :-
If X increases by 1 unit → Y changes by β₁
Hence slope interprets marginal effect.

Question 9
question :-
What is the purpose of the command:
pd.read_excel("[Link]")

options :-
a) Read Excel dataset into pandas DataFrame
b) Create new Excel file
c) Export regression results
d) Convert DataFrame into NumPy array

answer :-
a) Read Excel dataset into pandas DataFrame

theory :-
Pandas provides functions to import structured data into DataFrames.

solution / explanation :-
read_excel() loads Excel data → converts into DataFrame for analysis.

Question 10

question :-
If the 95% confidence interval for β₁ does not include 0, then:

options :-
a) Model is incorrect
b) Residuals are normal
c) Slope is statistically significant
d) Intercept is zero

answer :-
c) Slope is statistically significant

theory :-
If CI excludes 0 → null hypothesis (β₁ = 0) is rejected → predictor is significant.

solution / explanation :-
Since 0 not in interval → β₁ ≠ 0 → variable has significant effect.
Week 7

Question 1

question :-
When regression assumptions about the error term are violated, which of the following may occur?

options :-
a) Coefficient of determination becomes zero
b) Hypothesis testing results become unreliable
c) Regression line disappears
d) Sample size automatically reduces

answer :-
b) Hypothesis testing results become unreliable

theory :-
Regression assumptions (normality, homoscedasticity, independence) ensure valid statistical
inference. If violated, estimates may still exist, but inference (p-values, confidence intervals)
becomes unreliable.

solution / explanation :-
Violation affects hypothesis tests → leads to incorrect conclusions → hence results become
unreliable.

Question 2

question :-
Residual analysis is primarily used to:

options :-
a) Estimate regression coefficients
b) Test multicollinearity
c) Validate regression model assumptions
d) Maximize R²

answer :-
c) Validate regression model assumptions

theory :-
Residuals help check:

• Linearity

• Constant variance

• Normality

• Independence
solution / explanation :-
Residual plots reveal violations → used for validating assumptions, not for coefficient estimation.

Question 3

question :-
If the variance of residuals increases as the value of the independent variable increases, this
indicates:

options :-
a) Heteroscedasticity
b) Autocorrelation
c) Multicollinearity
d) Normality

answer :-
a) Heteroscedasticity

theory :-
Heteroscedasticity = non-constant variance of errors.

solution / explanation :-
Increasing spread of residuals → variance not constant → heteroscedasticity.

Question 4

question :-
A residual plot showing a clear curved pattern suggests:

options :-
a) The model fits perfectly
b) Constant variance exists
c) A linear model may be inappropriate
d) Errors are normally distributed

answer :-
c) A linear model may be inappropriate

theory :-
Residuals should be randomly scattered. A pattern (curve) indicates model misspecification.

solution / explanation :-
Curvature → relationship is non-linear → linear model is not suitable.

Question 5

question :-
Standardized residuals are primarily used to:
options :-
a) Detect outliers
b) Increase R²
c) Reduce bias in coefficients
d) Transform dependent variables

answer :-
a) Detect outliers

theory :-
Standardized residuals scale residuals → helps identify extreme observations.

solution / explanation :-
Values beyond ±2 or ±3 → indicate outliers.

Question 6

question :-
If residuals exhibit non-constant variance, the immediate consequence is:

options :-
a) Biased regression coefficients
b) Invalid hypothesis tests and confidence intervals
c) Incorrect sign of slope
d) Perfect multicollinearity

answer :-
b) Invalid hypothesis tests and confidence intervals

theory :-
Heteroscedasticity affects variance estimates → distorts standard errors.

solution / explanation :-
Incorrect standard errors → invalid t-tests and confidence intervals.

Question 7

question :-
In a multiple linear regression model, multicollinearity primarily affects:

options :-
a) The unbiasedness of regression coefficients
b) The magnitude of the dependent variable
c) The stability and standard errors of coefficient estimates
d) The calculation of residuals

answer :-
c) The stability and standard errors of coefficient estimates
theory :-
Multicollinearity → high correlation among predictors → unstable coefficients and inflated standard
errors.

solution / explanation :-
Leads to unreliable coefficient estimates → difficulty in interpretation.

Question 8

question :-
In multiple regression, the adjusted R² is preferred over R² because it:

options :-
a) Always increases when a new variable is added
b) Penalizes the inclusion of irrelevant independent variables
c) Eliminates multicollinearity
d) Guarantees better prediction accuracy

answer :-
b) Penalizes the inclusion of irrelevant independent variables

theory :-
Adjusted R² accounts for number of predictors and sample size.

solution / explanation :-
Unlike R², it decreases if irrelevant variables are added → better model selection metric.

Question 9

question :-
In a multiple regression model with k independent variables, the overall F-test is used to test whether:

options :-
a) All regression coefficients are individually significant
b) At least one independent variable is statistically significant
c) The intercept is equal to zero
d) Residuals are normally distributed

answer :-
b) At least one independent variable is statistically significant

theory :-
F-test checks:
H₀: β₁ = β₂ = … = βₖ = 0

solution / explanation :-
Reject H₀ → at least one predictor affects dependent variable.
Question 10

question :-
In a regression model with a dummy variable representing gender (Male = 1, Female = 0), the
coefficient of the dummy variable represents:

options :-
a) The average value of the dependent variable for males
b) The difference in mean dependent variable between males and females
c) The slope of the continuous independent variable
d) The variance of the dependent variable

answer :-
b) The difference in mean dependent variable between males and females

theory :-
Dummy variable coefficients measure change relative to base category.

solution / explanation :-
Female (0) is baseline → coefficient shows how much male differs from female.
WEEK 8

Question 1

question :-
In regression modeling, the key difference between linear regression and logistic regression is that:

options :-
a) Both require binary dependent variables
b) Linear regression requires continuous dependent variable while logistic regression requires binary
dependent variable
c) Logistic regression requires continuous dependent variable
d) Both require normally distributed dependent variables

answer :-
b) Linear regression requires continuous dependent variable while logistic regression requires binary
dependent variable

theory :-

• Linear regression → predicts continuous outcomes

• Logistic regression → predicts probabilities for categorical (binary) outcomes

solution / explanation :-
Linear → output is numeric
Logistic → output is probability (0–1) → used for classification

Question 2

question :-
Which measure is primarily used to assess model fit in logistic regression instead of the sum of
squared errors used in linear regression?

options :-
a) Adjusted R²
b) Mean absolute error
c) −2 Log Likelihood
d) Mean squared error

answer :-
c) −2 Log Likelihood

theory :-
Logistic regression uses likelihood-based estimation instead of minimizing squared errors.

solution / explanation :-
Model fit is evaluated using likelihood → commonly −2LL (deviance).

Question 3
question :-
Which statistical test is used to evaluate the overall significance of the logistic regression model?

options :-
a) F test
b) t test
c) G test
d) z test

answer :-
c) G test

theory :-
G-test (Likelihood Ratio Test) compares full model vs null model.

solution / explanation :-
Uses difference in log-likelihood → checks if model improves prediction.

Question 4

question :-
The coefficients in logistic regression are interpreted primarily using:

options :-
a) Standardized beta coefficients
b) Correlation coefficients
c) Variance inflation factors
d) Odds ratios

answer :-
d) Odds ratios

theory :-
Logistic coefficients represent log-odds; exponentiating gives odds ratios.

solution / explanation :-
exp(β) → tells multiplicative change in odds.

Question 5

question :-
In logistic regression, the odds ratio for an independent variable measures:

options :-
a) Change in odds for one-unit increase in predictor
b) Change in probability for one-unit increase
c) Change in mean response
d) Goodness of fit of the model
answer :-
a) Change in odds for one-unit increase in predictor

theory :-
Odds ratio = exp(β) → effect of predictor on odds.

solution / explanation :-
If OR > 1 → increase in odds
If OR < 1 → decrease in odds

Question 6

question :-
If the odds of an event occurring are 3, the corresponding probability is:

options :-
a) 0.5
b) 0.6
c) 0.75
d) 0.8

answer :-
c) 0.75

theory :-
Relation:
Probability = Odds / (1 + Odds)

solution / explanation :-
P = 3 / (1 + 3) = 3 / 4 = 0.75

Question 7

question :-
In logistic regression model evaluation, the difference between −2 log likelihood of the base model
and the proposed model follows approximately which distribution?

options :-
a) Chi-square distribution
b) t distribution
c) Normal distribution
d) F distribution

answer :-
a) Chi-square distribution

theory :-
Likelihood Ratio Test statistic follows chi-square distribution.
solution / explanation :-
Used to compare nested models → significance testing.

Question 8

question :-
The Wald test statistic used in logistic regression for testing significance of coefficients is:

options :-
a) β / SE(β)
b) β² / SE(β)
c) SE(β) / β
d) β / SE(β)²

answer :-
a) β / SE(β)

theory :-
Wald statistic evaluates significance of individual coefficients.

solution / explanation :-
Test statistic:
Z = β / SE(β)
Used to compute p-value.

Question 9

question :-
In logistic regression, the relationship between predictors and probability is modeled using:

options :-
a) Linear probability function
b) Exponential function
c) Logistic (sigmoid) function
d) Quadratic function

answer :-
c) Logistic (sigmoid) function

theory :-
Logistic function maps values to range (0,1).

solution / explanation :-
Ensures predicted probabilities remain valid.

Question 10
question :-
If the estimated probability of an event is P = 0.40, the corresponding odds are:

options :-
a) 0.4
b) 0.67
c) 1.5
d) 2.5

answer :-
b) 0.67

theory :-
Odds = P / (1 − P)

solution / explanation :-
Odds = 0.40 / 0.60 = 0.67
WEEK 9

Question 1

question :-
In a binary classification confusion matrix, which component represents records where the model
predicts class “1” but the actual class is “0”?

options :-
a) False Positive
b) True Positive
c) False Negative
d) True Negative

answer :-
a) False Positive

theory :-
Confusion matrix components:

• TP → predicted 1, actual 1

• FP → predicted 1, actual 0

• FN → predicted 0, actual 1

• TN → predicted 0, actual 0

solution / explanation :-
Predicted positive but actually negative → False Positive.

Question 2

question :-
Which of the following correctly defines classification accuracy?

options :-
a) (TP + FP) / Total
b) (TP + TN) / Total
c) (FP + FN) / Total
d) TP / (TP + FN)

answer :-
b) (TP + TN) / Total

theory :-
Accuracy = proportion of correct predictions.

solution / explanation :-
Correct predictions = TP + TN → divide by total observations.
Question 3

question :-
Which metric measures the proportion of actual positives correctly identified?

options :-
a) Precision
b) Recall (Sensitivity)
c) Specificity
d) Accuracy

answer :-
b) Recall (Sensitivity)

theory :-
Recall = TP / (TP + FN)

solution / explanation :-
Measures how many actual positives are captured.

Question 4

question :-
Reducing the cutoff value from 0.50 to 0.30 will generally:

options :-
a) Increase false negatives
b) Decrease sensitivity
c) Increase specificity
d) Increase predicted positives

answer :-
d) Increase predicted positives

theory :-
Lower threshold → easier to classify as positive.

solution / explanation :-
More observations labeled positive → predicted positives increase.

Question 5

question :-
In an ROC curve, the x-axis represents:

options :-
a) Sensitivity
b) Precision
c) False Positive Rate
d) Accuracy

answer :-
c) False Positive Rate

theory :-
ROC:

• X-axis → FPR = FP / (FP + TN)

• Y-axis → TPR (Recall)

solution / explanation :-
ROC plots trade-off between sensitivity and false positives.

Question 6

question :-
Which value of AUC (Area Under Curve) represents a perfect classifier?

options :-
a) 0.0
b) 0.5
c) 0.75
d) 1.0

answer :-
d) 1.0

theory :-
AUC measures separability:

• 1 → perfect

• 0.5 → random

solution / explanation :-
Perfect classifier distinguishes all classes correctly → AUC = 1.

Question 7

question :-
Which metric is most useful when false negatives are very costly (e.g., disease detection)?

options :-
a) Accuracy
b) Specificity
c) Sensitivity
d) Misclassification rate
answer :-
c) Sensitivity

theory :-
Sensitivity minimizes false negatives.

solution / explanation :-
In critical cases (like disease), missing a positive is dangerous → maximize recall.

Question 8

question :-
The first-order regression model with one predictor variable is represented as:

options :-
a) y = β₀ + β₁x₁ + ε
b) y = β₀ + β₁x₁ + β₂x₂ + ε
c) y = β₁x₁ + ε
d) y = β₁x₁² + ε

answer :-
a) y = β₀ + β₁x₁ + ε

theory :-
First-order model → linear relationship with intercept.

solution / explanation :-
Includes intercept and one predictor.

Question 9

question :-
In a second-order regression model with one predictor variable, which additional term is included?

options :-
a) x
b) x²
c) y²
d) xy

answer :-
b) x²

theory :-
Second-order model introduces curvature using squared term.

solution / explanation :-
Model: y = β₀ + β₁x + β₂x² + ε
Question 10

question :-
In regression analysis, an interaction term between two variables x₁ and x₂ is represented as:

options :-
a) x₁ + x₂
b) x₁ − x₂
c) x₁x₂
d) x₁ / x₂

answer :-
c) x₁x₂

theory :-
Interaction term captures combined effect of variables.

solution / explanation :-
Product term (x₁x₂) represents interaction.
WEEK 10

Question 1

question :-
A researcher wants to test whether region (categorical variable) and investment type (categorical
variable) are related. Which test is most appropriate?

options :-
a) Chi-square test of independence
b) t-test
c) ANOVA
d) z-test

answer :-
a) Chi-square test of independence

theory :-
Chi-square test is used to check association between two categorical variables.

solution / explanation :-
Both variables are categorical → use chi-square test of independence.

Question 2

question :-
Degrees of freedom in a contingency table with 4 rows and 3 columns:

options :-
a) 6
b) 12
c) 5
d) 7

answer :-
a) 6

theory :-
Formula:
df = (r − 1)(c − 1)

solution / explanation :-
df = (4−1)(3−1) = 3 × 2 = 6

Question 3
question :-
A Chi-square test shows significance, but several expected frequencies are below 5. What is the most
appropriate action?

options :-
a) Ignore the issue
b) Combine categories
c) Increase significance level
d) Use regression

answer :-
b) Combine categories

theory :-
Chi-square assumption: expected frequency ≥ 5.

solution / explanation :-
Low expected counts → merge categories to satisfy assumption.

Question 4

question :-
In a contingency table, if row and column variables are independent, then:

options :-
a) Observed = Expected
b) Observed > Expected
c) Observed < Expected
d) Cannot be compared

answer :-
a) Observed = Expected

theory :-
Under independence, expected frequencies match observed (in theory).

solution / explanation :-
Deviation between observed and expected indicates dependence.

Question 5

question :-
Which situation violates Chi-square assumptions?

options :-
a) Categorical data
b) Independent observations
c) Expected frequency < 5
d) Large sample size
answer :-
c) Expected frequency < 5

theory :-
Chi-square requires adequate expected counts.

solution / explanation :-
Small expected values → unreliable test results.

Question 6

question :-
Cluster analysis may give misleading results when:

options :-
a) Data is standardized
b) Variables are correlated
c) Variables are on different scales
d) Sample size is large

answer :-
c) Variables are on different scales

theory :-
Distance-based methods are sensitive to scale differences.

solution / explanation :-
Unscaled variables distort distance → incorrect clustering.

Question 7

question :-
Which method is most sensitive to outliers in clustering?

options :-
a) Hierarchical clustering
b) K-means clustering
c) Chi-square test
d) Regression

answer :-
b) K-means clustering

theory :-
K-means uses mean → highly affected by extreme values.

solution / explanation :-
Outliers shift centroids → unstable clusters.
Question 8

question :-
A marketer uses clustering to segment customers but finds unstable clusters. The most likely issue is:

options :-
a) Too many observations
b) Poor variable selection
c) High significance level
d) Low degrees of freedom

answer :-
b) Poor variable selection

theory :-
Relevant features are crucial for meaningful clustering.

solution / explanation :-
Irrelevant/noisy variables → inconsistent clusters.

Question 9

question :-
Which situation best suits cluster analysis?

options :-
a) Predicting sales
b) Grouping customers based on behavior
c) Testing independence
d) Estimating mean

answer :-
b) Grouping customers based on behavior

theory :-
Clustering is unsupervised learning → grouping similar data points.

solution / explanation :-
Used for segmentation tasks.

Question 10

question :-
Standardization transforms data so that:

options :-
a) Mean = 0 and standard deviation = 1
b) Mean = 1
c) Variance = 0
d) Values increase

answer :-
a) Mean = 0 and standard deviation = 1

theory :-
Standardization (Z-score):
Z = (X − μ) / σ

solution / explanation :-
Centers data at 0 and scales to unit variance.
WEEK 11

Question 1

question :-
In cluster analysis, dissimilarity between two objects is:

options :-
a) Always negative
b) Always zero
c) Non-negative and increases with difference
d) Equal to correlation

answer :-
c) Non-negative and increases with difference

theory :-
Dissimilarity measures (distance metrics) are always ≥ 0 and increase as objects become more
different.

solution / explanation :-
Distance cannot be negative → larger distance = more dissimilarity.

Question 2

question :-
If all values of a variable are missing, what should be done?

options :-
a) Replace with mean
b) Ignore missing values
c) Remove variable
d) Normalize

answer :-
c) Remove variable

theory :-
A variable with all missing values provides no information.

solution / explanation :-
Cannot impute meaningfully → best to remove it.

Question 3

question :-
Categorical variable dissimilarity is based on:

options :-
a) Ratio of mismatches
b) Mean difference
c) Variance
d) Correlation

answer :-
a) Ratio of mismatches

theory :-
For categorical data, dissimilarity is computed as proportion of unequal values.

solution / explanation :-
Mismatch count / total variables → dissimilarity.

Question 4

question :-
If two objects match perfectly in a categorical variable, dissimilarity is:

options :-
a) 1.0
b) 0.0
c) −1
d) Undefined

answer :-
b) 0.0

theory :-
Perfect match → no difference → zero dissimilarity.

solution / explanation :-
No mismatch → distance = 0.

Question 5

question :-
Why are ordinal variables standardized to [0,1]?

options :-
a) Increase variance
b) Reduce sample size
c) Improve correlation
d) Different variable scales need normalization

answer :-
d) Different variable scales need normalization

theory :-
Standardization ensures all variables contribute equally in distance calculations.
solution / explanation :-
Without scaling → variables with larger ranges dominate.

Question 6

question :-
If max = 3.08 and min = 1.34, normalization denominator is:

options :-
a) 1.74
b) 2.08
c) 3.08
d) 1.0

answer :-
a) 1.74

theory :-
Normalization uses: (max − min)

solution / explanation :-
3.08 − 1.34 = 1.74

Question 7

question :-
The objective function in K-means minimizes:

options :-
a) Between-cluster distance
b) Sum of squared distances within clusters
c) Correlation
d) Variance between clusters

answer :-
b) Sum of squared distances within clusters

theory :-
K-means minimizes within-cluster variance.

solution / explanation :-
Objective: minimize Σ||x − centroid||²

Question 8

question :-
Two objects are described by 4 categorical variables. They match in 3 variables and differ in 1 variable.
What is the dissimilarity?
options :-
a) 0.25
b) 0.5
c) 0.75
d) 1.0

answer :-
a) 0.25

theory :-
Dissimilarity = mismatches / total variables

solution / explanation :-
1 / 4 = 0.25

Question 9

question :-
Given log-transformed values: Object 1 = 2.65, Object 2 = 1.34, Maximum = 3.08, Minimum = 1.34.
Find normalized dissimilarity:

options :-
a) 0.5
b) 0.75
c) 1.0
d) 0.25

answer :-
b) 0.75

theory :-
Normalized distance:
|x₁ − x₂| / (max − min)

solution / explanation :-
|2.65 − 1.34| = 1.31
Range = 3.08 − 1.34 = 1.74
Dissimilarity = 1.31 / 1.74 ≈ 0.75

Question 10

question :-
Consider two centroids: C₁ = (1,1), C₂ = (5,5). A point P = (2,3). Which cluster will P belong to (using
Euclidean distance)?

options :-
a) Cluster 1
b) Cluster 2
c) Both equally
d) Cannot determine

answer :-
a) Cluster 1

theory :-
Assign to nearest centroid using Euclidean distance.

solution / explanation :-
Distance to C₁ = √[(2−1)² + (3−1)²] = √5
Distance to C₂ = √[(2−5)² + (3−5)²] = √13
√5 < √13 → closer to Cluster 1
WEEK 12

Question 1

question :-
In decision tree algorithm, attribute selection method is used to:

options :-
a) Clean data
b) Choose best splitting attribute
c) Remove outliers
d) Normalize features

answer :-
b) Choose best splitting attribute

theory :-
Attribute selection measures (Information Gain, Gini, Gain Ratio) help select the best feature for
splitting.

solution / explanation :-
Decision trees split data based on feature that gives best separation.

Question 2

question :-
CART follows which approach while building trees?

options :-
a) Top-down greedy
b) Bottom-up greedy
c) Random search
d) Backtracking

answer :-
a) Top-down greedy

theory :-
CART builds tree recursively by choosing best split at each step.

solution / explanation :-
Starts from root → splits downward → greedy decisions at each node.

Question 3

question :-
The Gini index is mainly used for:

options :-
a) Clustering
b) Classification
c) Regression only
d) Sampling

answer :-
b) Classification

theory :-
Gini index measures impurity in classification tasks.

solution / explanation :-
Lower Gini → purer node → better split.

Question 4

question :-
Information gain is biased towards:

options :-
a) Attributes with many values
b) Attributes with fewer values
c) Continuous attributes only
d) Binary attributes only

answer :-
a) Attributes with many values

theory :-
Information Gain favors attributes with many distinct values.

solution / explanation :-
Leads to overfitting → corrected using Gain Ratio.

Question 5

question :-
Gain ratio is used to:

options :-
a) Increase bias
b) Normalize dataset
c) Reduce dataset size
d) Remove bias of information gain

answer :-
d) Remove bias of information gain

theory :-
Gain Ratio adjusts Information Gain by intrinsic information.
solution / explanation :-
Penalizes attributes with many values.

Question 6

question :-
Which of the following ensures binary splits in decision trees?

options :-
a) Information gain
b) Gain ratio
c) Gini index
d) Entropy

answer :-
c) Gini index

theory :-
CART algorithm (uses Gini) produces binary splits.

solution / explanation :-
Each node splits into exactly two branches.

Question 7

question :-
In hierarchical clustering, HAC stands for:

options :-
a) Hierarchical Analytical Clustering
b) Hierarchical Agglomerative Clustering
c) Hybrid Agglomerative Clustering
d) High Accuracy Clustering

answer :-
b) Hierarchical Agglomerative Clustering

theory :-
HAC builds clusters by merging smaller clusters.

solution / explanation :-
Bottom-up clustering approach.

Question 8

question :-
The Euclidean distance formula is used to:
options :-
a) Measure similarity
b) Normalize data
c) Measure dissimilarity
d) Reduce dimensions

answer :-
c) Measure dissimilarity

theory :-
Distance metrics quantify how different two points are.

solution / explanation :-
Higher distance → more dissimilar.

Question 9

question :-
LabelEncoder in Python is used for:

options :-
a) Scaling data
b) Encoding categorical variables
c) Feature selection
d) Data splitting

answer :-
b) Encoding categorical variables

theory :-
LabelEncoder converts categories into numeric labels.

solution / explanation :-
Example: Male → 0, Female → 1.

Question 10

question :-
The function fit_transform() in encoding:

options :-
a) Only fits model
b) Only transforms data
c) Fits and transforms data
d) Deletes missing values

answer :-
c) Fits and transforms data
theory :-
fit() learns parameters, transform() applies them.

solution / explanation :-
fit_transform() combines both steps in one call.