Introduction to Statistics for Machine Learning

Definition of Statistics

Statistics is the science of collecting, organising, analysing, interpreting, and presenting data.
It provides the foundation for data-driven decision making — which is essential in machine learning.

Types of Statistics

There are two main types of statistics:

1. Descriptive Statistics
2. Inferential Statistics

1. Descriptive Statistics

Descriptive statistics summarise and organise data so that it can be easily understood.
They describe the basic features of a dataset, providing simple summaries about the sample and measures.

It involves:

• Measures of Central Tendency – e.g. mean, median, mode
• Measures of Dispersion (Variability) – e.g. range, variance, standard deviation
• Data Distribution – e.g. histograms, box plots, violin plots
• Summary Statistics – e.g. minimum, maximum, quartiles, percentiles

Example descriptive questions:

• What is the average height in a class?
• What is the most commonly purchased product in a store?
• What is the spread (variability) of exam scores in a class?

2. Inferential Statistics

Inferential statistics involve methods for making predictions, inferences, or generalisations about a population based on a sample of data.
They allow us to test hypotheses and estimate population parameters.

It involves:

• Hypothesis Testing
• P-Values and Confidence Intervals
• Statistical Tests, such as:
• • Z-Test
  • T-Test
  • Chi-Square Test
  • ANOVA (Analysis of Variance)

Example inferential questions:

• Are the average heights in this class similar to the average heights in other schools?
• Do customers in different stores tend to buy the same products?
• Does a new drug significantly improve recovery rates compared to an existing one?

Machine Learning Context

In machine learning, statistics helps in:

• Understanding data distributions before modelling (EDA – Exploratory Data Analysis)
• Feature scaling and normalisation
• Evaluating model performance using statistical tests
• Avoiding overfitting through sampling and hypothesis validation