Professional Skills
Each term below is a defined one. Select any chip to see the definition it is being used under, and where that definition comes from.
Research Interests
Areas I work in, each with the definition the field actually uses.
Data quality is the degree to which the characteristics of data satisfy stated and implied needs when used under specified conditions. The standard names the characteristics (accuracy, completeness, consistency, credibility, currentness among them) but does not prescribe how to measure them; supplying defensible measures is the methodological problem.
ISO/IEC 25012:2008, Data quality model
The branch of statistics concerned with methods whose use depends on substantial computation rather than closed-form results: simulation, resampling, Monte Carlo integration, numerical optimisation and iterative estimation.
Standard disciplinary usage
The study of algorithms that infer predictive models from data, analysed with the tools of statistical inference: estimation, the bias–variance decomposition, generalisation error and consistency, rather than predictive accuracy alone.
Standard disciplinary usage
The use of controlled experiments, case studies, field observation and measurement to obtain evidence about software artefacts, processes and the people who use them, instead of relying on argument from design.
Empirical software engineering, standard usage
The production of tested, documented and distributable implementations of statistical methods, such that a published result can be reproduced by a third party from the artefact alone.
Research software engineering practice
Statistical Methods
The methods themselves, defined before they are claimed.
Regression analysis estimates the relationship between a response variable and one or more explanatory variables, yielding a model of how the expected value of the response changes with those variables.
ISO 3534-1, Statistics — Vocabulary and symbols
Missing data is classified by the mechanism that produced it: missing completely at random, missing at random, or missing not at random, and the mechanism determines which treatments remain valid. An outlier is a member of a small subset of observations that appears inconsistent with the remainder of the sample.
Rubin (1976) for the mechanisms; ISO 16269-4 for outliers
Design of experiments is the arrangement in which an experimental programme is to be conducted, including the choice of factors and their levels, randomisation and blocking. An A/B test is the two-arm case: units are randomised between a control and one treatment.
ISO 3534-3, Design of experiments
Multivariate analysis treats observations on several variables jointly rather than one at a time, so that the dependence between variables is part of the model. The high-dimensional case is where the number of variables is comparable to, or exceeds, the number of observations.
Standard disciplinary usage
A research design that collects and integrates both numeric measurement and non-numeric evidence, with the integration made explicit at the design, collection or interpretation stage rather than left to the reader.
Standard methodological usage
Tools
What I build with, and what each one actually is.
A language and environment for statistical computing and graphics. My package work targets its distribution system, CRAN, whose policy defines what a publishable package must satisfy.
R Project / CRAN Repository Policy
A general-purpose programming language; here used with NumPy, pandas and scikit-learn for data handling and modelling.
Python Software Foundation
A Python web framework for building HTTP APIs, with request and response validation derived from type annotations.
Project documentation
A framework for building native mobile applications for iOS and Android from a single JavaScript or TypeScript codebase.
Project documentation
An SDK for compiled cross-platform applications, written in Dart, rendering its own widgets rather than wrapping platform controls.
Project documentation
A superset of JavaScript that adds static types, checked before the code runs and erased at compile time.
Project documentation
A platform for packaging an application together with its dependencies into an image that runs identically wherever the runtime is available.
OCI image specification
A distributed version control system in which every clone holds the full history, making the record of change auditable rather than centrally held.
Git documentation
An in-memory data structure store, used as a cache, message broker and short-lived database.
Project documentation
A hosted application platform providing authentication, a document database, storage and analytics.
Google Firebase documentation
An architectural style for networked applications built on stateless requests over uniform resource identifiers, using the HTTP method to express intent.
Fielding (2000), architectural style
Google Analytics 4, an event-based product analytics platform, where every interaction is recorded as a named event with parameters rather than as a pageview.
Google Analytics documentation