headways data quality?

[hamima-halim]

the scheduled headways in LAMP are pretty irregular. in gobble, we disregard the reported scheduled headways entirely and recalculate our own from gtfs.

• gtfs does not include NONREV-* or ADDED* stops
• gobble (realtime API) does report NONREV-* and ADDED* events
• LAMP does report NONREV-* and ADDED* events

THAT BEING SAID....
the headways we calculate seem to be identical the ones we report in gobble, and if we recalculate the lamp events in the same was we get the same numbers. but also, i might be illiterate.

[devinmatte]

We also process the monthly data files: https://github.com/transitmatters/t-performance-dash/blob/main/server/rapid/process_events.py

There may be some logic in here worth looking at (looks like we don't touch gtfs in this, so maybe not)

[devinmatte]

Aha I knew there was something: https://github.com/transitmatters/t-performance-dash/blob/main/server/bus/bus2train.py

We do this for bus monthly data we ingest (but for some reason not for train data, may want to update that)

[devinmatte]

We should look into smoothing headways into 15 or 30 minute averages to avoid these jumps in the bar