Day One, Thursday, July 8th
Arrive 12 noon - Arrival and registration in time for lunch.
2.15-3.45 Exposition and discussion of ‘the nature of classification'.
This session developed the argument that classification matters and that it is in fact fundamental to the investigation of causality when we are dealing with complex open systems. It drew on the arguments presented in Byrne (2002) Chapters Two and Six. The central argument, which is without doubt contentious and acceptance of which is not compulsory, is that our measurements do not describe real entities in the form of variables but rather traces of complex systems. These traces are in effect the product of the interaction of our measurement processes with the trajectories of complex systems at one or more given points in time. Numerical taxonomy procedures enable us to establish typologies according to criteria determined by us and to assign cases to the categories in those typologies. Note the emphasis on trajectories and we . We paid particular attention to time ordered classifications as an approach.
We concentrated here on basic hierarchical clustering using SPSS although students were informed about more sophisticated procedures, including fuzzy clustering. The general principle of these workshops is that understanding the underlying logic of case based enquiry is fundamental and that actual techniques are best acquired through personal practice - in summary once you know why you are doing what you are doing - then go away and do it. However, we introduced participants to the practical use of the basic methods and as much advanced stuff as we could get in. A particular emphasis was placed on the transmission of tricks of the trade and we welcomed additions to this repertoire of real practice from anyone in the group who themselves had experience of using the methods.
3.45 - 4.00 Tea break
4.00 - 6.00 - Exposition and Demonstration of numerical taxonomy techniques
This session took the form of a demonstration of methods by us in which we tried to engage with as many clustering techniques as possible. We showed both the basics of the approaches and our bag(s) of tricks which are concerned with enabling sensible interpretation of output. The intention was to set people up to work on their own with at least two data sets - one using continuous data and one using binarized categorical data - in the subsequent workshops.
If people had data sets they wanted to bring and work on themselves that was fine - ideally as an SPSS export data file although EXCEL files were fine. There is no real problem with inputting continuous data of the form used for any regression based procedure. Two-step cluster available in SPSS 12 allows for multinomial input of categorical and ordinal data so the old laborious business of binarizing is no longer necessary and we can use typical survey output data as well.
Ordinal data - well if bold enough input as continuous - otherwise binarize - or even do both.
6.15 - 7 - Evening meal
7.15 - 9 First Hands On workshop
You work yourselves with demonstrator support.
Day Two, Friday, July 9th
9 - 10.45 Second Hands On workshop As last night.
10.45 - 11.00 Coffee break
11.00 - 12.30 Reflective Presentations
Three short presentations by people reflecting on how the approaches might work in relation to their own research problems.
12.30 - 1.30 Lunch
1.30 - 3.30 Reflective plenary discussion - small group based with feedback
Us doing qualitative research on you - where participants met in focus groups of around eight and reviewed their thinking on case based methods in general and clustering procedures in particular.
3.30 Tea and depart