talking data

Day 2 of the master class was a round table discussion with people from a variety of areas human–computer interaction, bio-mechanics, software development, mathematics and urban regeneration.

Several participants introduced a particular topics or probe  they are facing with their data and this led to discussions that connected to a wide variety of areas.

Topics we discussed included:

  • building explanatory models from questionnaire data
  • dealing with relatively small data sets with relatively large numbers of measured values for each data element
  • data reduction – principle components and factor analysis
  • doing easy understandable analysis for exploring data even if not the perfect model/method for the final publishable analysis
  • stratified samples (re-weighting analysis/results based on population statistics)
  • deliberately having unbalanced samples that emphasise a particular group
  • dangers of spurious results from large volumes of data, Bonferroni correction and other techniques
  • SPSS and use of the Laerd resources – I’d not come across this, but highly recommended by participants
  • trying multiple models until one ‘works’
  • how to make sense of some of the complex models possible in tools such as SPSS, R, etc.
  • combining model-driven and data-driven analysis
  • making use of large-volume poor-fidelity data with small-volume high-fidelity data
  • dealing with one-off data where it is hard or impossible to gather more
  • Montey-Hall problem

I will endeavour to write short blogs about each of these ave time … do chase me if there is one that you’d particularly like me to priorities!