EMET8002 Case Studies in Applied Economic Analysis and Econometrics

How EMET8002 is assessed

The six tasks total 100%, as listed in the course introduction. The presentation is marked out of 10 for presentation skills, 10 for the research question and hypothesis, 10 for data and methods and 20 for preliminary results, discussion and questions. In the 2026 offering the report deadline was extended to midnight on 1 June because of a public holiday. Two announcements give different closing dates for the take-home assignment, 15 May and 15 June 2026, so check the assignment page on Canvas. Extensions and extenuating circumstances applications go through the ANU Request for assessment adjustment app.

Current EMET8002 dates

Dates are as published in the Semester 1, 2026 course announcements and the Week 10 presentation page on the course site. Confirm exact deadlines and submission settings in the live LMS.

What EMET8002 covers

Thirteen chapters follow the lectures, from survey data and regression through categorical, time series and panel models to causal designs, prediction and the research report.

Its lectures cover the main methods an empirical project relies on: survey data and hypothesis tests, linear regression and the assumptions behind it, models for binary, multinomial, ordered and count outcomes, time series with trends, unit roots, lag models and volatility, panel data with fixed and random effects, and the causal designs of difference-in-differences, regression discontinuity and instrumental variables.

An additional lecture introduces machine learning for prediction.

How it is taught

Lectures run twice a week for the first seven weeks, and weekly computer labs apply each method in Stata to datasets such as an ANU Poll survey, US state education data, PISA 2015, World Bank GDP series and a wage panel.

From Week 8 the lecture slots become consultation sessions for the project, and Week 10 is set aside for student presentations.

What the assessment rewards

Most of the grade attaches to one research project.

You choose a published paper from an approved list, propose an extension by Week 6, present your progress in Week 10, and submit a report of 4,000 to 5,000 words that replicates the paper and then extends it with new data, a different method or a different population. Problem sets, lab attendance and a take-home assignment of questions and data analysis make up the rest.

The course rewards transparent explanations of every data and modelling decision, and it does not allow AI or translation tools. This guide gives each block of lectures its own chapter, with original worked examples whose numbers are computed in full.

Worked example · free

Estimating a programme effect with difference-in-differences

Key terms

Null Hypothesis

The default claim of no effect or no difference that a test assumes true until the evidence rejects it.

Ordinary Least Squares

The estimator that chooses coefficients to minimise the sum of squared residuals.

Endogenous Regressor

An explanatory variable that is correlated with the error term, which biases its estimated coefficient.

Omitted Variable Bias

Bias that arises when a left-out variable both determines the outcome and correlates with an included regressor.

Marginal Effect

The change in a predicted probability when one regressor moves from a stated starting value.

Odds Ratio

The odds of an event in one group divided by the odds in another group.

Incidence Rate Ratio

The multiplier on an expected count that a one-unit change in a regressor produces in a Poisson model.

Unit Root

A stochastic trend in a time series, present when the coefficient on its own lag equals one.

Spurious Regression

A regression in which trending series look related even though they are not.

Fixed Effects

A panel estimator that gives each unit its own intercept to absorb all of its time-invariant characteristics.

Parallel Trends

The assumption that, absent the treatment, the treated group's outcome would have moved in step with the control group's.

Running Variable

The observed score whose crossing of a threshold determines treatment in a discontinuity design.

Exclusion Restriction

The requirement that an instrument affects the outcome only through the endogenous regressor.

EMET8002 FAQ

Questions students ask in each EMET8002 chapter

What is the difference between a Type I and a Type II error?

A Type I error rejects a null hypothesis that is actually true, like convicting an innocent defendant, and its probability is the significance level. A Type II error fails to reject a false null, like acquitting a guilty defendant; one minus its probability is the power of the test. Chapter 01 · Survey Data, Sampling and Hypothesis Tests

What does holding other factors constant mean in a multiple regression?

Each slope measures the change in the outcome for a one-unit change in its own regressor while every other included regressor stays fixed. It does not hold fixed anything left out of the model, which is why omitted variables remain a concern. Chapter 02 · Simple and Multiple Linear Regression

Does multicollinearity bias regression coefficients?

No. It inflates the variance of the affected coefficients, so their partial effects become imprecise and may look insignificant. The lecture's checks are pairwise correlations beyond 0.8, or 0.7 in some texts, and a variance inflation factor above 10. Chapter 03 · When Regression Assumptions Fail

Why not just use OLS for a binary outcome?

OLS gives the linear probability model, which is easy to interpret but has heteroskedastic, non-normal errors and can predict probabilities outside the range from 0 to 1. Logit and probit avoid impossible predictions by using a curved link function. Chapter 04 · Binary Outcomes With Logit and Probit

What is the independence of irrelevant alternatives assumption?

It says a person's choice between two options is unaffected by what other options are available. The red bus and blue bus example shows how it fails when a new option is a close substitute for an existing one, which is when a nested or mixed logit is safer. Chapter 05 · Multinomial, Ordered and Count Outcomes

What is a spurious regression?

It is a regression in which two series appear strongly related only because both contain trends. The lecture's example is US GDP and federal debt, significant together in levels but not in first differences, and the Week 6 lab repeats the point with Australian population and US GDP. Chapter 06 · Time Series Data, Trends and Autocorrelation

Why can't a unit root be tested with an ordinary t-test?

Under the null of a unit root, the test statistic does not follow the usual distribution, so the standard critical values are wrong. The Dickey-Fuller test uses its own critical values, which statistical tables give and Stata calculates. Chapter 07 · Unit Roots, Lag Models and Volatility

When should I use fixed effects instead of random effects?

Use fixed effects when the unobserved unit characteristic is likely correlated with the regressors, which a Hausman rejection supports. Prefer random effects when that correlation is implausible, when units change little over time, or when you need to estimate a time-invariant variable. Chapter 08 · Panel Data With Fixed and Random Effects

What is the parallel trends assumption?

It assumes that, had the treatment never happened, the treated group's outcome would have changed by the same amount as the control group's. It cannot be tested directly, because the counterfactual is unobserved, but parallel pre-treatment trends in an event study support it. Chapter 09 · Difference-in-Differences and Event Studies

What is the difference between a sharp and a fuzzy RDD?

In a sharp design, treatment jumps from 0 to 1 at the cutoff, so the outcome jump is the effect. In a fuzzy design, only the probability of treatment jumps, so the outcome jump is divided by the treatment jump, which is estimated by two-stage least squares. Chapter 10 · Regression Discontinuity and Kink Designs

What makes an instrument valid?

It must be relevant, meaning correlated with the endogenous regressor, which a first-stage F-test can check, and exogenous, meaning uncorrelated with the error so that it affects the outcome only through the regressor. The second condition cannot be tested and must be argued. Chapter 11 · Instrumental Variables and Two-Stage Least Squares

What is the bias-variance trade-off?

Simple, high-bias models miss real patterns and underfit, while complex, high-variance models fit the training data closely but generalise badly. The best predictive model balances the two, which is why it is chosen by its error on data it has not seen. Chapter 12 · Machine Learning for Economic Prediction

How long should the research report be?

The course introduction sets 4,000 to 5,000 words. The report has two parts, a replication of the chosen paper's main findings and an extension of it, and the Week 7 lecture proposes a structure from introduction and literature through data, methods and both sets of results to a conclusion. Chapter 13 · Presenting Results and Writing the Research Report

How to prepare for the assessments

Work backwards from the research report. In the first two weeks, read the approved papers, pick one whose data you can actually obtain, and start any HILDA or Compustat application at once, because access can take weeks and every student must apply individually. Then use the labs as rehearsals: when a lab covers the method your paper uses, run the same commands on your own data that week.

Before the proposal, write the research question as a pair of hypotheses and name the assumption your extension depends on, such as parallel trends or an exclusion restriction. While replicating, check your observation counts and descriptive statistics against the paper before comparing coefficients, and keep a log of every cleaning decision, since explaining those choices is what the course rewards.

Prepare the presentation as a progress report with no more than three results slides, and rehearse answers to likely questions. Leave the final week for tables, interpretation sentences with units, and a conclusion that states the contribution.

Your AI Economics tutor for EMET8002

Stuck on a hard EMET8002 question? Sia is AskSia’s AI Economics tutor — ask any EMET8002 Case Studies in Applied Economic Analysis and Econometrics question and get a clear, step-by-step explanation grounded in how the course is actually taught and assessed. Read this whole study guide free, then take your hardest questions to Sia.

STAT7038 Regression Modelling · STAT7055 Introductory Statistics for Business and Finance · ECON7074 Macroeconomic Analysis