This is a list of current tasks under research and or development.
The different status values are:
- Research - preliminary investigations underway looking at approaches, algorithms and checking of maths, software packages etc.
- Experimental - the work is being undertaken standalone producing output for specific projects, often in paper (not web) form. These are normally associated with new ideas.
- Workbench Development - the work is being integrated with the workbench.
Goodness of Fit tests with cobweb diagrams
In this work, we can mix dyads, triads, stones and MCQs into I x J contingency tables and test goodness of fit - are the relationships as we'd expect them to be.
Triads by Cohort
In this work we look at the user providing cohort pairs or triplets and we produce triad graph output side by side showing the means and 1 variance from the mean. This then becomes a visual comparison to check similarity between the cohort groups.
Examples of cohort pairs would be 1. Older verses younger respondents, 2. Male verses female, 3. Those who respond positively verses those who respond negatively.
Status: Research with some Experimental
Older versions of Explorer enabled a user to generate correlation cobweb diagrams of correlations between triad, dyad and stone anchors.
We are not entirely comfortable with this approach so we are experimenting with correlations between triad/dyad/stone zones. The basic idea a bootstrap on the signifiers replicating independence and then a chi-squared test on the null hypothesis of the data being independent compared to the bootstrap - we graph the squared residuals.
By focussing on zones this does mean that we can mix triads, dyads and stones.
A number of different graphic options are available. We are currently focussing on cobweb diagrams - the current version a .ps file but will need to be updated to be more web friendly for the Workbench.
Tests of triad means across cohorts
Related to the triads by cohort - this is a significance test version based on non-parametric t tests. Given a cohort pair, are the differences in means statistically significant.
Status: Workbench Development
Linling the 2.4 vocabulary into the Workbench so that the user can dynamically select the language (and later the vocabulary).
===X,Y and X,Y Contour maps
Status: Workbench Development
Being able to dynamically specify triad/dyad/stone anchors and generate an x,y plot with contour map (and head map) options.
Also provide for a cross product of triad and dyad and stone graphs.
NOTE: x,y plots of compositional data needs to be treated with care. This will be considered during the development. IN particular, one should not mix triad and dyad and stone values in the same plot.
Main current workpiece for this: Establish in Shiny how to enable brush on dynamically generated output.
Decision Trees for Triad (Dyad/Stone) Zones
Based on the work of Anna Hancher, introducing decisions trees to evaluate the MCQ breakdown within triad zones (to extend to dyads and stones).
It responds to the question of how significant the various MCQs are to the make up of a triad zone.