What is Band A about?

Band A is about data in context. Modern data science is unusual in that data is being collected, often without a question in mind. This is a change from classical statistics and science, where data is normally collected to answer a particular question.

Band A is about the usability of data. In the early days, one could argue that the traditional statistical pipeline contained only Band A.

The Sense of Band A

  • In statistics it roughly maps to confirmatory data analysis
  • In Band A we consider the appropriateness of a given data set to answer a particular question, or to be subject to a particular analysis.

Often in Band A we might find our existing data (although it might be at an advanced level of readiness, e.g. B1) is not sufficient to answer the questions we have. Band A may require the acquisition of new data sets or further work in annotation to make the data usable. Any of those other required data sets may need to go through the full process again.

Whereas Band B might involve characterising the missing values, Band A could involve modelling them.

At the end of Band A the data set is ready for RAMP, Kaggle, OpenML, etc