To be awarded the AQM qualifier, students need undertake an empirical, quantitative Capstone Research project. It is essential that the methods used demonstrate the student’s ability to analyse quantitative data and interpret the results in a competent way.
There are many ways to achieve this which will vary with the discipline, however in general there are three components that should be present:
- Descriptive statistics
- Inference
- Multivariate models
Most students should be encouraged to use existing datasets. An exception is where a randomised experiment is proposed and a credible plan for recruiting participants can be demonstrated.
Benchmarks, guidelines and examples:
-
Secondary analysis of existing numerical datasets – including but not limited to survey data, administrative records such as social security payment records, educational attainment records, health records, court records, parliamentary records, macroeconomic or socio-political indicators or other population level datasets
- Collection and primary analysis of data from surveys or randomised experiments, for instance survey-based experiments using MTurk, Prolific or other online platforms
- Data from lab experiments
- Statistical principles and techniques
Capstone projects should demonstrate the appropriate use of:
-
Basic descriptive statistics (for example means and percentages)
- Inferential methods, either frequentist or Bayesian (for example p-values, confidence or credible intervals, hypothesis testing, standard errors, posterior density)
- Some form of multivariate modelling (for example multiple OLS regression, logit or probit analysis)
- Survival analysis, econometrics, fixed and random effects models, time series analysis, factor analysis and structural equation models
- Machine learning methods
Where randomised experiments form the empirical data for a project, multivariate analyses may not be so necessary, although covariate adjustment and other exploratory analyses could be employed to demonstrate ability to carry out and interpret multivariate techniques.