anonymised feature output to facilitate sharing

[andreaskutka]

Problem
Very little data is available for testing or training models. Often users are constrained of sharing their survey data for other use.

Solution:
Develop an intermediate output that contains the features in a raw anonymised form that can be shared more easily by users. Ideally, the output is prior to score transformation, to allow potential use as training data. The output should:

• replace the variable names and responsible names with letters
• remove answer option, question text, etc, only keep columns needed
• save the feature tables on item and unit level as csv, and zip them