- GLO home
- Adaptive statistical methods
- Expert seminar
- Workshops
- links
- contact
- Initial research on implementation
- Links to statistical websites
Initial research on implementation
This document contains research information and thoughts about how the implementation of a website to generate statistics teaching RLOs might work.
Systems for statistical analysis
R is a popular GNU statistics package based on the commercial package S. It is heavily used in the statistics laboratory in Cambridge and most other academic institutions. R can be invoked from the command line and can export plots in PNG format, almost ideal for embedding in web pages, apart from a lack of support in Internet Explorer until quite recently. See http://www.r-project.org/
Zope is a popular python-based object-publishing environment. It may be ideal for this project. There is already a project, RStatServer, to enable R and Zope to be integrated through SOAP, although this project is at a relatively early stage. However, the product Rpy, which enables R to be called directly from python, could be used; the advantage of RSOAP is that it is thread-safe.
To display equations, a Zope product exists at http://products.bluedynamics.org/Products/LatexMethod/ which converts Latex into images.
Future progress
As this project progresses, there are some important points to keep in mind. The first is that our primary aim should be to make this tool as useful as possible for students, not as easy to use as possible for staff, although this is clearly also very important if it is to be utilised at all.
Before many decisions can be made, some idea is required of exactly what it is that the tool will be used to teach. This will provide information as to how standardised or flexible templates needs to be. Decisions will need to be made regarding how large a single RLO is. Here there are a few possible approaches:
- An RLO for teaching a complex idea will make you work through all the ideas required to understand it beforehand, similarly to the STEPS material.
- RLOs will require you to know all the background information before you start, requiring you to have read RLOs if you don’t have enough knowledge.
- RLOs will have ‘asides’ which you can follow if you don’t know about a required concept. Whether these are also interactive is an independent question.
The RLOs should probably reference a glossary. This should be pitched at a level such that it explains all the concepts used comprehensively, but does not confuse by going on to more advanced topics. Whether this level can be agreed between subjects must be decided. For extended reading, the glossary could reference a more advanced site such as Mathworld. The glossary will probably have to point out the differences in terminology and sometimes in definitions between subjects.
It should be noted at an early stage that there are many different concepts to be conveyed in teaching statistics, of which the easiest to learn and to teach are how to do statistical tests or to draw graphs. Understanding when a test is relevant and how to choose which test to use is much more complex, and harder to teach.
Hypothesis testing is perhaps the hardest to understand and important thing in statistics, and is brushed over by the STEPS material. It is an example of something that is very much subject-independent, and thus fits the template model (if indeed anything needs to be changed between subjects at all) very well, though a lot of work needs to go into designing the template.
System consideration
From a system point of view, one thing that needs to be decided is whether teaching staff can access, copy, or edit other staff’s RLOs. Are all RLOs going to be viewable to all students? How will staff direct students to learning material? Will the system log whether they visit it? Who is responsible for the glossary?
Maybe one strength of this system is that staff should be able to select a number of topics, and the system will be present the student with a combined RLO for those topics.
If staff could build on and develop other staff’s resources then over time a very useful repository of teaching resources may grow.
A facility for students to rate the usefulness of an RLO and to comment on what it doesn’t explain well would be a useful feature.
Implementation details
Although these are not important, there are a few things to keep in mind. One is that equations are going to need to be entered. I believe that making graphics from Latex source is currently still the best way of displaying mathematics on the web. Maybe MathML will make it in the end, but it isn’t there yet.
If questions are going to be part of the teaching site, deciding how they affect the flow of the site is a significant design decision..
