Xml has:
    id
    route tag
    dir tag
    lat
    lon
    secsSinceReport
    predictable
    heading

scrapetimetable.py:
    Get route data, including lat and lon for stops
    Get timetable eg schedule for stops

findspacings.py:
    Actually calculate the spacing between arrivals at a stop
    Also calculates arrival time at each stop
    Needs the h5 files to already be generated
    Saved in data.mat

analyze_all_named_stops.py:
    uses findspacings.py
    generates spacings.mat
    has:
    data_map = {
            'stop_idxs': [x[0] for x in all_data],
            'schedule_codes': [x[1] for x in all_data],
            'chunk_idxs': [x[2] for x in all_data],
            'times'     : [x[3] for x in all_data],
            'spacings': [x[4] for x in all_data],
            'spacings_expected': [x[5] for x in all_data],
            }

plot.py:
    Uses spacing.mat to generate plots
    Just uses stop ids and spacings
    stuff = scipy.io.loadmat('spacings.mat')
    stop_idxs = stuff['stop_idxs'].transpose()
    all_gaps = stuff['spacings'].transpose()

dependencies:
    gcc
    hdf5
    scipy
    tables
    bs4


I already have times, but I need expected times, not just expected spacings.
I can get expected time for a stop if it is on the schedule, but what about
stops which do not have a schedule?  What to do then?


plots.jl
work on finding dwell time
finite accel/veloc/etc
refine gps locations
instantaneous speed as proxy for traffic


julia con
intro to julia
david sanders
https://www.cs.ubc.ca/~murphyk/MLbook/
principal components analysis
springerlink
9.520, 6.860
9.6: generalized linear mixed models