Enhance MTR graphing utility functions

[martinholmer]

This pull request builds on from the substantial contribution made by @GoFroggyRun in merged pull request #948. Read the two mtr_graph_* function docstrings for details on how this plotting capability works now.

Most of the changes rationalize exiting features or add new capabilities. But the logic of MARS filtering option was changed so that the the income percentiles are constructed after the filtering (not before).

I'm interested in receiving comments and suggestions on how this can be further improved.

@MattHJensen @feenberg @Amy-Xu @GoFroggyRun @andersonfrailey @codykallen @zrisher

[codecov-io]

Current coverage is 98.48% (diff: 100%)

Merging #1020 into master will increase coverage by 0.27%

@@             master      #1020   diff @@
==========================================
  Files            35         35          
  Lines          2180       2243    +63   
  Methods           0          0          
  Messages          0          0          
  Branches          0          0          
==========================================
+ Hits           2141       2209    +68   
+ Misses           39         34     -5   
  Partials          0          0          

Powered by Codecov. Last update c8ce60f...023eb74

[codykallen]

mtr_plot uses two lines of deprecated code. Lines 984 and 985 have:
fig.legend.legend_spacing = 5
fig.legend.legend_padding = 5

However, the latest documentation indicates that this should be coded as:
fig.legend.spacing = 5
fig.legend.padding = 5

Source here, at the bottom of the page
: http://bokeh.pydata.org/en/latest/docs/user_guide/styling.html#legends

[martinholmer]

@codykallen said:

mtr_graph_plot uses two lines of deprecated code. Lines 984 and 985 have:
fig.legend.legend_spacing = 5
fig.legend.legend_padding = 5

However, the latest documentation indicates that this should be coded as:
fig.legend.spacing = 5
fig.legend.padding = 5

Thanks for pointing this out, @codykallen. I found I was using an older version of bokeh. Now that I've upgraded to version 0.12.3 (and added that requirement to the environment.yml file) and changed the mtr_graph_plot code, these warning messages no longer appear.

[martinholmer]

I've generated and looked at quite a few mtr graphs and now think the work on this pull request is complete. But I'm hoping others will use the pull request code and the puf.csv file to generate other graphs to see if they make sense.

In order to help in that effort, I'm posting here a script that generates 18 different versions of the e00200p MTR graph for a pop-the-cap social security payroll tax reform. I can't figure out how to convert the resulting HTML file (which contains all 18 graphs) into one PNG file that I can post here. So, I'll post after the script just two of the graphs that were puzzling to me. Do you think they make sense?

First, the following script is in the mtr-graphs.py file located in top directory of the Tax-Calculator code tree along with the puf.csv file and is invoked like this: python mtr-graphs.py.

from taxcalc import *
import bokeh.layouts as blayout
ref = {2013: {'_SS_Earnings_c': [1e99]}}  # "pop-the-cap" reform in 2013
bp.output_file('mtr-graphs.html', title='MTR by Income Percentile')
gplots = list()
for inc in ['wages', 'agi', 'expanded_income']:
    for mtr in ['ptax', 'itax', 'combined']:
        for dwt in [False, True]:
            calc1 = Calculator(policy=Policy(), records=Records())
            calc2 = Calculator(policy=Policy(), records=Records())
            calc2.policy.implement_reform(ref)
            gdata = mtr_graph_data(calc1, calc2,
                                   mars='ALL',
                                   mtr_measure=mtr,
                                   mtr_wrt_full_compen=False,
                                   income_measure=inc,
                                   dollar_weighting=dwt)
            gplot = mtr_graph_plot(gdata)
            gplots.append(gplot)
bp.show(blayout.column(gplots))

Second, here are two of the 18 graphs generated by the above script.

@MattHJensen @feenberg @Amy-Xu @GoFroggyRun @andersonfrailey @codykallen @zrisher

[martinholmer]

If there are no concerns about pull request #1020, it will be merged at end of the Nov 1st workday.

@MattHJensen @feenberg @Amy-Xu @GoFroggyRun @andersonfrailey @codykallen @talumbau

[MattHJensen]

Thanks a lot @martinholmer. I expect that #1021 means there is no longer a change in MTR for IIT when you pop the cap as is shown in the second graph above. Do you have any ideas at this point about what is causing the blips for payroll tax MTR for the 15-20 percentiles of AGI in the first chart?