Web Services

Status: API Complete, Prototyping Started, Requirements Needed

User Stories

As a researcher (EDM, M&M, CMDM), I want to retrieve transaction data programmatically and append custom fields to transactions in DataShop so that I can more easily run my models on the data outside of DataShop and put the results back in.
As a researcher with a machine learning background, I want to retrieve student-step data programmatically so that I can more easily run my analyses on the data without much human intervention.

As a researcher who has created a custom field for an existing data set, I want to import that new field into the step rollup table so that it is both preserved and available in DataShop tools. - Ryan Baker, 12/15/2008

Notes/Comments

• Adding custom fields to existing datasets at Transaction Level only
• Needed for M&M and CMDM clusters [Ryan Baker, October 2008]
• Will enable researchers to use output from existing models
  • Gaming Detector [Arnon Hershkovitz, Mihaela Cocea, Summer 2008]
  • Bayesian Knowledge Tracing. [Hao Cen, migrated from feature wish list]
• Key to collaboration.
• Ability to automatically add columns to DataShop data sets at transaction level [Ryan, Startup Memo, summer 2008]
• Points:
  • Tx level, Need user incentive to upload values, need associated notes, need graphing of

Semi-structured interviews with researchers

• Ido Roll: main conclusions are that being able to use the new columns in the same ways one can use DataShop-native colums might be crucial; data is often not at transaction level (eg, student variable), so DS might do the tx-level mapping for the user based on key column(s)
  
• Phil Pavlik, email: "I guess I might use the adding of a column by itself if the opportunity was there to look at it as a model graph in DataShop. Often it is good to visualize the results of a model."

• Get Step Rollup
  • How to show the KCs, multiple columns? one column? use labels or 1's and 0's? -- Ken Koedinger, DataShop Team Meeting, Aug 28, 2009
    • Multimapped - column for every KC, then 1 or 0 (this is the Q matrix structure)
    • The other way to go, is to have one column for all the KCs

---

See complete DataShop Feature Wish List.