Project Meeting 2024.12.17

Agenda

• Estimation Enhancements

Technical

• Estimation Enhancements
  • Functionality enhancements are fully implemented
    • Old code has been removed
    • File type format is now parquet as default
    • Added "Estimation_table_recipes"
    • Tested on SANDAG example
  • Larch enhancements
    • Added Analyze Predictions tool
      • Similar to Alogit's predict function; Takes choices and model probabilities of the choices and segment by attribute of the choosers
        • Calculates statistical significance between chooser attributes
        • Shows results as a boxplot-esque figure.
        • Can customize bins for each chooser attributes
        • Can apply a weighting function, to reweight observations to count for relative frequencies in your dataset vs actual
          • For MNL estimation, won't use weights; but for viewing against segmented attributes you can review using the weights
      • Added Elasticity calculation
        • calculates the elasticity of demand with respect to various input variables
        • Can segment by population or chooser attribute
      • Added Full Probability Array
        • Can change anything and recalculate the probability array
  • Next steps - finish clean-up and testing; complete documentation; test/implement with PSRC (post-TRB, end of January)