Introduction

What Statistics Is, and What This Course Covers

Course Overview

  • Statistical inference and decision theory
  • Point estimation: methods and properties
  • Interval estimation and confidence intervals
  • Hypothesis testing: theory and applications
  • Asymptotic theory and limit theorems
  • Bayesian inference

The thread running through all of it

We observe a sample. We want to say something about the population it came from — and quantify how sure we are.

Statistics

Sampling goes one way; the interesting (and hard) direction is the other one.

Statistics

  • Statistics can generally be divided into two parts:
    • Descriptive statistics — summarising the data you actually have
    • Inferential statistics — drawing conclusions beyond the data you have
  • Statistics typically involves analysing a sample taken from a population.
  • From the sample, we generally make statements about the population.
  • It is therefore important that the sample is representative of the population.

Working assumption

For most of this course we will simply assume that the samples are representative. When that assumption fails, no amount of clever mathematics repairs it.

Sampling, Live

Draw a sample from a known population and watch how well \(\bar{x}\) recovers \(\mu\).

What the Simulation Shows

  • With random sampling, larger \(n\) shrinks the error — and the error is centred on zero: press “draw” repeatedly and it wanders either side of \(\mu\).
  • With a biased scheme, larger \(n\) does not help. The error stays stubbornly on one side, no matter how much data you collect.

The point

Sample size fixes noise. It never fixes bias. Everything in this course assumes the sampling itself was done honestly.

Where We Go From Here

Next: Random Variables

  • Turning outcomes into numbers
  • CDFs, PMFs, PDFs
  • Moments and common distributions
  • Laws of large numbers, CLT

Then: Inference

  • Multivariate random variables
  • Estimation theory
  • Hypothesis testing