Statistical Advisor

Project Proposal by Martin Stacey


Statistical Advisor

Software

An object oriented programming language such as Smalltalk or Java, or possibly Prolog or Lisp or an expert systems shell

Covers

Artificial intelligence, system design and programming, statistics

Skills Required

Systems analysis and design, programming, preferably some interest in artificial intelligence, not afraid of maths, basic knowledge of statistics, some interest in how science gets done

Challenge

Conceptual ???? Technical ?? Programming ??

Brief Description

In many sciences, including biology, psychology and sociology, researchers need to know how to find out whether they should really believe what their numbers appear to be telling them. They also need to know how to design tests that will give them sets of numbers from which they can get real results. Software developers run into the same problems when they want to run experiments to measure the characteristics of their user interfaces. Statistics is the branch of mathematics concerned with finding out what numerical data really means. There is a variety of statistical calculations that can be applied to different types of data set, and which rest on different assumptions about the nature of the data. People who want to get answers from experiments need to understand statistical arguments, but they don't need to be highly expert in statistics - they can (and should) get advice on experimental design and analysis from the experts.

Your mission in this project is to develop a software system that can advise users on what experimental design to employ, and which statistical tests to apply to their data. The system needs to ask the users questions about the nature of their problems, and apply some decision procedures for proposing appropriate designs and tests. The project involves a significant HCI challenge in finding a good way for the users to interact with the system so that they understand the questions it is asking and can provide usable answers.


Back to