Project Meeting 2016.05.20

Large Scale Performance Test

• CDAP
  • Jeff (@toliwaga) implemented HH level chunking for the CDAP model
  • CDAP chunking is working, but CDAP model fails on full population; Jeff investigating issue
• NAs
  • Some choice models are resulting in all alternatives not available (see #81).
  • Jeff temporarily fixed this by allowing the model to choose an alternative, even though all alternatives are NA
  • We need to log and exit the simulation when this actually occurs so we can track down the data/expression errors
• Logging/tracing
  • Wu (@wusun2) pointed out that we also need to remove NA alternatives from the choice set before calculating probabilities (see #81).
  • We haven't started working on logging, but we should, especially since we now need tracing to identify the bad data/expressions noted above.
  • Tracing output needs to be text, easily parsed, and importable into Excel; see #92

Next Steps

• Continue to track down and fix issues for a full model run
• Implement logging and tracing of an individual household

Project Admin

• Fixing the issues in #81 is not really in the scope, but the project team agrees that our plan for moving forward is what we should be doing