rakeshkumar nayak

Posted on Oct 3

Descriptive vs Inferential Statistics: A Practical Guide with Real-World Examples....

#analytics #data #datascience

Descriptive and Inferential Statistics: From Data to Decisions

From Mean and Median to Probability, Hypothesis Testing, Confidence Intervals, Correlation, and Regression

Written by Rakesh Kumar Nayak

Statistics is one of the most important foundations of Data Science, Data Analytics, Machine Learning, and Artificial Intelligence.

Whenever we work with data, two fundamental questions arise:

What does the data tell us?
What can we conclude about a larger population from the data?

The first question is mainly answered by Descriptive Statistics, while the second is answered by Inferential Statistics.

In this article, we will go step-by-step from the basics of statistics to probability distributions, sampling, confidence intervals, hypothesis testing, correlation, and regression—with practical examples.

1. What Is Statistics?

Statistics is the science of collecting, organizing, analyzing, interpreting, and presenting data.

For example, suppose the salaries of five employees are:

₹25,000, ₹30,000, ₹35,000, ₹40,000, ₹50,000

We may want to know:

What is the average salary?
What is the middle salary?
How much do salaries vary?
What is the probability that a randomly selected employee earns more than ₹40,000?
Can we use a sample of employees to estimate the average salary of the entire company?

These questions take us into descriptive statistics, probability, and inferential statistics.

2. Population vs Sample

Before understanding inferential statistics, we need to understand two important concepts.

Population

A population is the complete group we are interested in.

Example:

All employees working in a company.

Sample

A sample is a smaller subset selected from the population.

Example:

100 employees selected from the company.

Studying the entire population is often expensive or time-consuming, so we commonly work with samples.

Example

Population = 10,000 employees

Sample = 500 employees

We calculate statistics from the 500 employees and use those results to learn about the larger population.

3. Descriptive Statistics

Descriptive statistics summarizes and describes the data that we already have.

The major areas include:

Measures of central tendency
Measures of dispersion
Measures of position
Distribution shape
Data visualization

4. Measures of Central Tendency

Central tendency tells us where the center of the data lies.

The three most common measures are:

Mean
Median
Mode

5. Mean

The mean is the arithmetic average.

Formula

Mean = Sum of all observations / Number of observations

Suppose five students scored:

60, 70, 80, 90, 100

Mean:

(60 + 70 + 80 + 90 + 100) / 5 = 80

Therefore, the average score is 80.

Python Example

import numpy as np

scores = [60, 70, 80, 90, 100]

print(np.mean(scores))

Output:

80.0

6. Median

The median is the middle value after sorting the data.

Example:

20, 30, 40, 50, 60

Median = 40

For an even number of observations:

20, 30, 40, 50

Median:

(30 + 40) / 2 = 35

Why is Median Important?

The median is less affected by extreme values.

Consider:

20, 25, 30, 35, 500

The mean is strongly affected by 500, while the median remains 30.

Therefore, the median is often useful for highly skewed data such as salaries, house prices, and income.

7. Mode

The mode is the most frequently occurring value.

Example:

10, 20, 20, 30, 40

Mode = 20

Mode is particularly useful for categorical data.

Example:

Red, Blue, Blue, Green, Blue

Mode = Blue

8. Mean vs Median vs Mode

Measure	Meaning	Common Use
Mean	Average	Numerical data without strong skew
Median	Middle value	Data containing outliers or skew
Mode	Most frequent value	Categorical or frequency-based data

9. Measures of Dispersion

Knowing the average is not enough.

Consider two datasets:

Dataset A:

48, 49, 50, 51, 52

Dataset B:

10, 30, 50, 70, 90

Both have a mean of 50.

However, their variability is very different.

This is why we need measures of dispersion.

Important measures include:

Range
Variance
Standard deviation
Interquartile range

10. Range

Range measures the difference between the maximum and minimum values.

Formula

Range = Maximum - Minimum

For:

10, 20, 30, 40, 50

Range:

50 - 10 = 40

11. Variance

Variance measures how far observations are spread around the mean.

Population variance:

σ² = Σ(X - μ)² / N

Sample variance:

s² = Σ(X - x̄)² / (n - 1)

A larger variance indicates greater variability.

12. Standard Deviation

Standard deviation is the square root of variance.

σ = √Variance

Standard deviation is one of the most commonly used measures of variability.

Python Example

import numpy as np

data = [10, 20, 30, 40, 50]

print("Mean:", np.mean(data))
print("Variance:", np.var(data))
print("Standard Deviation:", np.std(data))

13. Understanding Distribution Through Visualization

A histogram helps us understand how numerical observations are distributed.

For example, an exam-score dataset can show:

Center
Spread
Skewness
Possible outliers
Overall distribution shape

Figure 1: Distribution of Exam Scores — Mean and Median

[Insert Figure 1 here]

14. Quartiles and Percentiles

Quartiles divide ordered data into four parts.

Q1

25th percentile

Q2

50th percentile, which is the median

Q3

75th percentile

The Interquartile Range is:

IQR = Q3 - Q1

IQR is particularly useful for identifying potential outliers.

15. Box Plot

A box plot visually represents:

Minimum
Q1
Median
Q3
Maximum
Potential outliers

Figure 2: Box Plot and Five-Number Summary

[Insert Figure 2 here]

Box plots are especially useful when comparing distributions across multiple groups.

16. Probability

Probability measures the likelihood that an event will occur.

Basic Formula

P(A) = Number of favorable outcomes / Total number of possible outcomes

For a fair coin:

P(Head) = 1 / 2 = 0.5

Therefore, the probability is 50%.

Probability always lies between:

0 ≤ P(A) ≤ 1

Where:

0 = Impossible

1 = Certain

17. Basic Probability Rules

Addition Rule

For mutually exclusive events:

P(A ∪ B) = P(A) + P(B)

Multiplication Rule

For independent events:

P(A ∩ B) = P(A) × P(B)

Complement Rule

P(Aᶜ) = 1 - P(A)

These rules form the foundation of probability calculations.

18. Conditional Probability

Conditional probability measures the probability of an event when another event has already occurred.

Formula

P(A|B) = P(A ∩ B) / P(B)

Example

Suppose we know that a customer purchased a laptop.

What is the probability that the same customer also purchased a laptop bag?

This is an example of conditional probability.

19. Bayes' Theorem

Bayes' theorem allows us to update the probability of an event after receiving new information.

Formula

P(A|B) = [P(B|A) × P(A)] / P(B)

Bayes' theorem is widely used in:

Spam detection
Fraud detection
Medical diagnosis
Recommendation systems
Classification
Naive Bayes algorithms

20. Random Variables

A random variable represents a numerical outcome of a random experiment.

For example:

Let X = number of heads obtained when tossing a coin three times.

Possible values:

0, 1, 2, 3

There are two major types.

Discrete Random Variable

Takes countable values.

Examples:

Number of customers
Number of calls
Number of defective products

Continuous Random Variable

Can take any value within a range.

Examples:

Height
Weight
Temperature
Time

21. Probability Distributions

A probability distribution describes how probabilities are assigned to possible outcomes.

Important probability distributions include:

Uniform Distribution
Bernoulli Distribution
Binomial Distribution
Poisson Distribution
Normal Distribution
Exponential Distribution

Understanding these distributions is important for Data Science and statistical modeling.

22. Uniform Distribution

In a uniform distribution, outcomes within a specified interval have equal probability density.

For example, a random number generated between 0 and 1 follows a continuous uniform distribution if every value in that interval is equally likely in the density sense.

23. Bernoulli Distribution

Bernoulli distribution represents an experiment with exactly two possible outcomes.

Examples:

Success / Failure
Yes / No
Pass / Fail
1 / 0

If:

P(Success) = p

Then:

P(Failure) = 1 - p

Bernoulli distribution is the foundation of the binomial distribution.

24. Binomial Distribution

Binomial distribution models the number of successes in a fixed number of independent Bernoulli trials.

Formula

P(X = k) = C(n,k) × pᵏ × (1-p)ⁿ⁻ᵏ

Example

A coin is tossed 10 times.

What is the probability of getting exactly 5 heads?

Here:

n = 10

p = 0.5

k = 5

Figure 3: Binomial Distribution

[Insert Binomial Distribution Figure here]

25. Poisson Distribution

Poisson distribution is commonly used to model the number of events occurring during a fixed interval of time or space.

Formula

P(X = k) = e⁻λ × λᵏ / k!

Examples:

Number of customer calls per hour
Number of website visitors per minute
Number of machine failures per month
Number of accidents per day

Figure 4: Poisson Distribution

[Insert Poisson Distribution Figure here]

26. Normal Distribution

The normal distribution is one of the most important distributions in statistics.

It has a characteristic bell-shaped curve.

It is defined by:

μ = Mean

σ = Standard Deviation

The standard normal distribution has:

μ = 0

σ = 1

Figure 5: Standard Normal Distribution

[Insert Normal Distribution Figure here]

27. The 68–95–99.7 Rule

For approximately normally distributed data:

Approximately 68% of observations lie within 1 standard deviation of the mean.

Approximately 95% lie within 2 standard deviations.

Approximately 99.7% lie within 3 standard deviations.

This is called the Empirical Rule.

28. Z-Score

A z-score tells us how many standard deviations an observation is from the mean.

Formula

Z = (X - μ) / σ

Example

Mean = 70

Standard deviation = 10

Student score = 90

Z = (90 - 70) / 10

Z = 2

Therefore, the score is 2 standard deviations above the mean.

29. Inferential Statistics

Now we move from describing data to making conclusions.

Inferential statistics uses sample data to estimate population parameters and test hypotheses.

Major concepts include:

Sampling
Sampling distributions
Central Limit Theorem
Point estimation
Confidence intervals
Hypothesis testing
p-values
t-tests
z-tests
ANOVA
Chi-square tests
Correlation
Regression

30. Sampling

Suppose a company has 50,000 customers.

Instead of surveying all 50,000 customers, we select 1,000 customers.

The 1,000 customers form our sample.

The objective is to use information from the sample to learn about the larger population.

Good sampling aims to reduce bias and obtain a sample that reasonably represents the target population.

31. Sampling Distribution

Suppose we repeatedly take samples from a population and calculate the mean of every sample.

The collection of those sample means forms the sampling distribution of the sample mean.

Figure 6: Sampling Distribution of the Sample Mean

[Insert Sampling Distribution Figure here]

This concept is fundamental to statistical inference.

32. Central Limit Theorem

The Central Limit Theorem is one of the most important concepts in statistics.

Under common conditions, as the sample size becomes sufficiently large, the sampling distribution of the sample mean approaches a normal distribution, even when the original population is not normally distributed.

This is one reason we can make statistical inferences using sample data.

33. Point Estimation

A point estimate provides one value as an estimate of a population parameter.

For example:

Sample mean = 72

We may use 72 as an estimate of the population mean.

However, a single number does not communicate the uncertainty around the estimate.

This leads to confidence intervals.

34. Confidence Interval

A confidence interval provides a range of plausible values for a population parameter.

A simplified confidence interval for a mean is:

Estimate ± Critical Value × Standard Error

For a 95% confidence interval, the commonly used standard normal critical value is approximately 1.96 when the normal approximation is appropriate.

Example

Suppose:

Sample mean = 72

Standard error = 2

Then:

72 ± 1.96 × 2

The interval is approximately:

68.08 to 75.92

Figure 7: 95% Confidence Interval

[Insert Confidence Interval Figure here]

Important Interpretation

A 95% confidence level does not mean that there is a 95% probability that a particular already-computed interval contains the fixed population parameter.

In frequentist statistics, it means that if we repeatedly constructed intervals using the same method, approximately 95% of those intervals would contain the true parameter.

35. Hypothesis Testing

Hypothesis testing provides a formal framework for evaluating claims about a population.

We generally define two hypotheses.

Null Hypothesis

H₀

The baseline assumption.

Alternative Hypothesis

H₁ or Hₐ

The competing hypothesis.

36. Hypothesis Testing Example

Suppose a company claims:

"Average delivery time is 30 minutes."

We collect a sample of delivery times.

We could formulate:

H₀: μ = 30

H₁: μ ≠ 30

We then select an appropriate statistical test, calculate the test statistic, and evaluate the evidence against the null hypothesis.

37. Significance Level

The significance level is represented by:

A commonly used value is:

α = 0.05

It defines the threshold used in the statistical decision rule.

38. P-Value

The p-value measures how compatible the observed data are with the null hypothesis under the assumptions of the statistical test.

A common decision rule is:

If p < α:

Reject H₀.

If p ≥ α:

Do not reject H₀.

Important

A p-value is NOT:

The probability that H₀ is true
The probability that the result occurred by chance
A measure of the size or practical importance of an effect

Statistical significance and practical significance are different concepts.

39. Type I and Type II Errors

Statistical decisions involve uncertainty.

Type I Error

Rejecting a true null hypothesis.

Its probability is commonly associated with:

Type II Error

Failing to reject a false null hypothesis.

Its probability is commonly represented by:

Statistical power is:

Power = 1 - β

40. Two-Tailed Hypothesis Test

A two-tailed test checks whether a parameter differs from a reference value in either direction.

For α = 0.05 under the standard normal distribution, the critical values are approximately:

-1.96

+1.96

Figure 8: Two-Tailed Hypothesis Test

[Insert Hypothesis Testing Figure here]

41. One-Tailed vs Two-Tailed Tests

One-Tailed Test

Used when the alternative hypothesis is directional.

Example:

H₁: μ > 50

Two-Tailed Test

Used when the alternative hypothesis is non-directional.

Example:

H₁: μ ≠ 50

The choice should be based on the research question and study design, ideally before looking at the results.

42. Common Statistical Tests

Statistical Test	Typical Application
One-Sample t-Test	Compare one sample mean with a reference value
Independent t-Test	Compare means of two independent groups
Paired t-Test	Compare two measurements from the same subjects
Z-Test	Mean/proportion testing under appropriate assumptions
ANOVA	Compare means across multiple groups
Chi-Square Test	Test association between categorical variables
Pearson Correlation	Measure linear association
Regression	Model relationships and make predictions

The correct test depends on the research question, data type, study design, and assumptions.

43. T-Test Example

Suppose we want to investigate whether the average salary of a sample differs from ₹40,000.

We can define:

H₀: μ = ₹40,000

H₁: μ ≠ ₹40,000

A one-sample t-test may be appropriate when the population standard deviation is unknown and the required assumptions are reasonably satisfied.

Python Example

from scipy.stats import ttest_1samp

salary = [38000, 42000, 41000, 39000, 45000]

stat, p_value = ttest_1samp(salary, 40000)

print("Test Statistic:", stat)
print("P-value:", p_value)

44. ANOVA

ANOVA stands for:

Analysis of Variance

It is commonly used to compare the means of three or more groups.

Example:

Suppose we want to compare average salaries across:

Data Analytics
Data Science
Software Engineering

We can formulate:

H₀: μ₁ = μ₂ = μ₃

The alternative hypothesis states that not all population means are equal.

If ANOVA provides sufficient evidence against H₀, further analysis may be required to identify which groups differ.

45. Chi-Square Test

The Chi-square test is commonly used with categorical variables.

Example:

Suppose we want to investigate whether:

Gender

and

Purchase Decision

are associated.

Example data:

Gender	Purchased	Not Purchased
Male	120	80
Female	140	60

A Chi-square test can be used to assess whether there is statistical evidence of an association between these categorical variables.

46. Correlation

Correlation measures the strength and direction of a linear relationship between two numerical variables.

Pearson correlation coefficient:

-1 ≤ r ≤ +1

Interpretation:

r = +1 → Perfect positive linear relationship

r = 0 → No linear relationship

r = -1 → Perfect negative linear relationship

For example, study hours and exam scores may show a positive correlation.

47. Correlation Does Not Mean Causation

This is one of the most important lessons in statistics.

Suppose ice cream sales increase when swimming pool attendance increases.

The two variables may be correlated.

However, this does not mean ice cream sales cause people to visit swimming pools.

A third variable, such as temperature, could influence both.

Therefore:

Correlation ≠ Causation

48. Regression

Regression is used to model relationships between variables and make predictions under appropriate assumptions.

For simple linear regression:

Y = β₀ + β₁X + ε

Where:

Y = Dependent Variable

X = Independent Variable

β₀ = Intercept

β₁ = Slope

ε = Error Term

Example

We can model exam score based on study hours.

Figure 9: Correlation and Simple Linear Regression

[Insert Correlation & Regression Figure here]

49. Descriptive vs Inferential Statistics

Descriptive Statistics	Inferential Statistics
Describes observed data	Draws conclusions about a population
Mean	Confidence Interval
Median	Hypothesis Testing
Mode	P-Value
Range	t-Test
Variance	ANOVA
Standard Deviation	Chi-Square Test
Percentiles	Regression
Charts	Population Estimation

A simple way to remember the difference:

Descriptive Statistics tells us what happened in the data.

Inferential Statistics helps us reason about what the data may imply beyond the observed sample.

50. Complete Statistical Workflow for a Data Science Project

A practical statistics workflow can look like this:

Step 1 — Define the Problem

Understand the business or research question.

Step 2 — Collect Data

Data may come from:

CSV files
Excel
Databases
APIs
Web scraping
Surveys

Step 3 — Clean the Data

Check for:

Missing values
Duplicates
Incorrect data types
Outliers
Inconsistent values

Step 4 — Perform Exploratory Data Analysis

Calculate:

Mean
Median
Mode
Range
Variance
Standard deviation
Quartiles
Percentiles

Step 5 — Visualize the Data

Use:

Histograms
Box plots
Bar charts
Scatter plots
Line charts
Distribution plots

Step 6 — Understand Probability

Identify appropriate probability concepts and distributions.

Step 7 — Formulate Hypotheses

Define:

H₀

and

H₁

Step 8 — Select the Statistical Test

Choose based on:

Data type
Number of groups
Study design
Distribution
Statistical assumptions

Step 9 — Calculate the Test Statistic

Apply the selected statistical method.

Step 10 — Interpret Results

Consider:

P-value
Confidence interval
Effect size
Practical significance

Step 11 — Communicate the Findings

Translate statistical results into meaningful business or research insights.

51. Practical Data Science Example

Imagine an e-commerce company wants to understand customer spending.

It collects data from 1,000 customers.

Variables include:

Customer_ID
Age
Gender
Income
Purchase_Amount
Products_Purchased
Rating

Descriptive Analysis

We calculate:

Average Purchase Amount
Median Purchase Amount
Standard Deviation
Minimum
Maximum
Q1
Q3

Suppose:

Mean Purchase Amount = ₹2,500

Median Purchase Amount = ₹2,200

Standard Deviation = ₹900

These statistics describe the observed sample.

52. Probability Analysis

The company may ask:

"What is the probability that a randomly selected customer spends more than ₹3,000?"

The answer could be estimated using empirical probabilities or an appropriate statistical model, depending on the data and assumptions.

53. Inferential Analysis

Suppose the company wants to compare average spending between two marketing campaigns.

We could formulate:

H₀: μA = μB

H₁: μA ≠ μB

An appropriate statistical test can then be selected based on the study design and assumptions.

The resulting confidence interval, p-value, and effect size can help describe the evidence.

54. Statistics in Machine Learning

Statistics is deeply connected with Machine Learning.

It helps us understand:

Data

Distribution, variability, missing values, and outliers.

Features

Relationships and associations between variables.

Model Evaluation

Performance measurements and uncertainty.

Sampling

Training and testing datasets.

Probability

Classification and probabilistic predictions.

Hypothesis Testing

Evaluating whether observed differences may be explained by random variation.

Regression

Understanding relationships and prediction.

55. Statistics in Generative AI and Modern Data Science

Even with modern Artificial Intelligence and Generative AI, statistical thinking remains important.

Data Scientists may use statistics for:

Data exploration
Experiment design
A/B testing
Model evaluation
Sampling
Uncertainty estimation
Error analysis
Feature analysis
Business decision-making

AI tools can generate code, but understanding the statistical reasoning behind that code remains important.

56. Most Important Statistical Formulas

Mean

x̄ = Σx / n

Population Variance

σ² = Σ(x - μ)² / N

Standard Deviation

σ = √σ²

Z-Score

Z = (X - μ) / σ

Conditional Probability

P(A|B) = P(A ∩ B) / P(B)

Bayes' Theorem

P(A|B) = P(B|A)P(A) / P(B)

Binomial Distribution

P(X=k) = C(n,k)pᵏ(1-p)ⁿ⁻ᵏ

Poisson Distribution

P(X=k) = e⁻λ λᵏ / k!

Confidence Interval

Estimate ± Critical Value × Standard Error

Simple Linear Regression

Y = β₀ + β₁X + ε

57. What Should a Data Scientist Remember?

You do not need to memorize every statistical formula without understanding it.

Focus on understanding:

Descriptive Statistics

Mean
Median
Mode
Variance
Standard deviation
Range
Quartiles
Percentiles
IQR
Skewness

Probability

Basic probability
Conditional probability
Bayes' theorem
Independent events
Random variables

Probability Distributions

Bernoulli
Binomial
Poisson
Normal
Uniform
Exponential

Inferential Statistics

Sampling
Sampling distribution
Central Limit Theorem
Confidence intervals
Hypothesis testing
P-values
Type I and Type II errors
Statistical power

Statistical Tests

t-test
z-test
ANOVA
Chi-square
Correlation
Regression

Most importantly, understand when to use a method, why you are using it, what assumptions it requires, and how to interpret the result.

58. Final Takeaway

Statistics is much more than calculating an average.

The complete journey can be remembered as:

Data → Descriptive Statistics → Probability → Sampling → Sampling Distribution → Estimation → Confidence Intervals → Hypothesis Testing → Correlation → Regression → Decision Making

Descriptive statistics helps us understand the data we have.

Probability helps us quantify uncertainty.

Inferential statistics helps us learn from samples and reason about populations.

For anyone preparing for a career in Data Analytics, Data Science, Machine Learning, or AI, statistics provides one of the strongest foundations for working with real-world data.

About the Author

Rakesh Kumar Nayak

I am passionate about transforming raw data into meaningful insights and continuously developing my skills in Data Science, Analytics, Machine Learning, and AI.

Written by Rakesh Kumar Nayak