More consistent treatment of accessibilities desired

[jfdman]

Is your feature request related to a problem? Please describe. When accessibilities are not defined consistently across model components, model sensitivities can be illogical. For example, in the testing of explicit error terms, we found that decreased transit costs led to longer work tour lengths. Because the telecommute frequency model uses distance to work instead of a mode choice logsum for accessibility, this resulted in lower work tours in the build alternative. MWCOG found a similar issue when testing the effect of increased parking cost on telecommute frequency.

Describe the solution you'd like I suggest that future estimation efforts attempt to use more consistent accessibility terms, even when those terms may not be as significant as less consistent terms like distance. The Oregon Joint Model Estimation Project is a good opportunity to do so, depending on which models get re-estimated.

Describe alternatives you've considered An alternative to estimation would be to attempt to convert existing distance terms (and potentially other terms) to logsums using simple relationships between distance and logsums to scale the existing distance coefficient, but this would be less than ideal due to potential errors in calculations, and some models may use ranges of distance rather than a linear distance term which adds complications.