Let's consider Chicago's historic black belt, West Garfield Park, East 
Garfield Park, North Lawndale, Near West Side, Near South Side, Douglas, 
Oakland, Grand Boulevard, Washington Park, and Englewood (see William Julius 
Wilson, When Work Disappears, p. 13). I found the corresponding community area numbers 
here: http://en.wikipedia.org/wiki/Community_areas_in_Chicago

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
                                       EXAMPLE 1:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

In these 10 neighborhoods, let's ask, which other types of calls predict calls about
vacant and abandoned buildings? Here's the command to enter:

python example-streams.py --area "[26,27,28,29,33,35,36,38,40,68]" --predict vacant --output results/black-belt-vacant.csv

And here are the results I get... (I'd love to know what results you get with a larger dataset!)

-------------------- In area [26.0, 27.0, 28.0, 29.0, 33.0, 35.0, 36.0, 38.0, 40.0, 68.0]
Correlation = 0.13107
for predicting vacant with these leading indicators: tree debris
# of events in X: 1352
# of events in Y: 2069
search complete, output written to results/black-belt-vacant.csv

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
                                       EXAMPLE 2:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Now let's ask, in which tracts of these neighborhoods does "tree debris" best predict vacant calls?

python example-locations.py --area "[26,27,28,29,33,35,36,38,40,68]" --predict vacant --streams ["tree debris"] --output results/black-belt-vacant-tracts.csv

Results:

-------------------- In area [26.0, 27.0, 28.0, 29.0, 33.0, 35.0, 36.0, 38.0, 40.0, 68.0]
Correlation = 0.63585
for predicting vacant with these leading indicators: tree debris
in these tracts [2808.0, 2804.0, 3510.0, 3515.0, 3805.0, 8329.0, 8333.0, 8381.0, 2809.0, 2828.0, 2819.0, 2838.0, 8436.0, 3511.0, 3802.0, 8414.0, 8371.0, 8410.0, 8358.0]
# of events in X: 175
# of events in Y: 72
search complete, output written to results/default-output.csv

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
                                       EXAMPLE 3:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Finally, what if we want to search over both tracts and streams simultaneously in these areas? This search takes
awhile...

python example-simultaneous.py --area "[26,27,28,29,33,35,36,38,40,68]" --predict vacant --output results/black-belt-vacant-simultaneous.csv

Results:

-------------------- In area [26.0, 27.0, 28.0, 29.0, 33.0, 35.0, 36.0, 38.0, 40.0, 68.0] with streams [garbage cart black maintenance graffiti removal occupied
 pot hole in street rodent baiting/rat complaint sanitation code violation
 street lights - all/out tree debris]
Correlation = 0.52265
for predicting vacant with these leading indicators: garbage cart black maintenance, graffiti removal, sanitation code violation
# of events in X: 1888
# of events in Y: 314
search complete, output written to results/default-output.csv


~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
                                       EXAMPLE 3:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
BONUS: now let's search over circles centered on each tract in the city, with radius .01 (this can be changed on line 3 of the
code---it'd be great if you could mess around and figure out a reasonable search radius)
For each circle, we'll look for a subset of tracts and streams where the correlation with vacant is high. 
To try to raise the level of significance, we'll do the search with data from the first half of the year, then
calculate the correlation for the second half of the year for the tracts/streams that were selected, and also for the
entire year. These are reported as R (entire year), R_a (first half), and R_b (second half). 

NB: this search takes awhile! Consider letting it run in the background while you do something else. If you use the "-u" option for
python then you can take a look at the results in another Terminal window by typing "tail -f results/vacant-centeredontracts-sig.csv"

python -u example-centeredontracts-sig.py --predict vacant --output results/vacant-centeredontracts-sig.csv


~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
                            And some more examples of syntax
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

python example-streams.py --tracts "[8430.0, 8431.0, 8432.0, 8433.0, 8434.0, 8435.0, 8436.0, 8437.0, 8438.0]"
python example-streams.py "pot hole in street" --area "[6]"

python example-locations.py --area "[34,35]" 
python example-locations.py --predict vacant

This one will take a long time to run; while it's running you can check
the results by opening another terminal and looking at results/vacant.csv 
(the -u option for python tells it to write results right away without 
waiting until the program exits.)

python -u example-centeredontracts.py --predict vacant --output results/vacant.csv