notes.txt

Number of distinct rds_index (route-direction-stop) (calls_2015-10): 25,262

date trips in 2015-10:  1,471,140

stop times in 2015-10: 56,485,803

join calls to calls to get observed headway
not necessary, retained here for testing

````
DROP TABLE IF EXISTS call_headways;
CREATE TABLE call_headways (
    call_id INTEGER NOT NULL PRIMARY KEY,
    headway MEDIUMINT UNSIGNED NOT NULL
);

INSERT INTO call_headways (call_id, headway)
SELECT
    a.`call_id`,
    (TIME_TO_SEC(TIMEDIFF(b.`call_time`, a.`call_time`)) - IF(a.`dwell_time` > 0, a.`dwell_time`, 0)) AS headway
FROM
    calls a
    LEFT JOIN call_increments c1 ON (c1.`call_id`=a.`call_id`)
    JOIN calls b ON (a.`rds_index`=b.`rds_index`)
    LEFT JOIN call_increments c2 ON (c2.`call_id`=b.`call_id`)
WHERE
    c1.`stop_increment` - 1  = c2.`stop_increment`
    AND b.`call_time` > a.`call_time`;
````

Phase 2:

* “Conservative” EWT, down to the stop level
    excess wait time, omitting any 'missing' buses from the analysis

    Observed/actual headway - scheduled headway, exclude calls where next bus is "missing"
    Already have A.H. and S.H., just need to find calls that are "missing"

* On-time terminal departures, at the route level
    percentage of buses in excess of three minutes behind schedule as of the third stop on each given route

* Wait assessment, down to the stop level
    calculated according to the MTA's definition for buses (something like, headways exceeding the schedule by 3 minutes or more for high frequency service; or 5 minutes or more for low frequency service)

* Stop spacing, down to the stop level
    how far apart are bus stops

* Circuitous routes, at the route level
    how indirect is the path of a given route relative to a direct (straight line) route

* Route-level EVT
    "excess in-vehicle time" (the difference between scheduled trip time and actual trip time from end-to-end of the route)