In this exercise, reproduce some of the findings from What Makes Houston the Next Great American City? | Travel | Smithsonian, specifically the calculation represented in
whose caption is
To assess the parity of the four major U.S. ethnic and racial groups, Rice University researchers used a scale called the Entropy Index. It ranges from 0 (a population has just one group) to 1 (all groups are equivalent). Edging New York for the most balanced diversity, Houston had an Entropy Index of 0.874 (orange bar).
The research report by Smithsonian Magazine is Houston Region Grows More Racially/Ethnically Diverse, With Small Declines in Segregation: A Joint Report Analyzing Census Data from 1990, 2000, and 2010 by the Kinder Institute for Urban Research & the Hobby Center for the Study of Texas.
In the report, you'll find the following quotes:
How does Houston’s racial/ethnic diversity compare to the racial/ethnic diversity of other large metropolitan areas? The Houston metropolitan area is the most racially/ethnically diverse.
....
Houston is one of the most racially/ethnically diverse metropolitan areas in the nation as well. *It is the most diverse of the 10 largest U.S. metropolitan areas.* [emphasis mine] Unlike the other large metropolitan areas, all four major racial/ethnic groups have substantial representation in Houston with Latinos and Anglos occupying roughly equal shares of the population.
....
Houston has the highest entropy score of the 10 largest metropolitan areas, 0.874. New York is a close second with a score of 0.872.
....
Your task is:
Tabulate all the metropolian/micropolitan statistical areas. Remember that you have to group various entities that show up separately in the Census API but which belong to the same area. You should find 942 metropolitan/micropolitan statistical areas in the 2010 Census.
Calculate the normalized Shannon index (entropy5) using the categories of White, Black, Hispanic, Asian, and Other as outlined in the Day_07_G_Calculating_Diversity notebook
Calculate the normalized Shannon index (entropy4) by not considering the Other category. In other words, assume that the the total population is the sum of White, Black, Hispanic, and Asian.
Figure out how exactly the entropy score was calculated in the report from Rice University. Since you'll find that the entropy score reported matches neither entropy5 nor entropy4, you'll need to play around with the entropy calculation to figure how to use 4 categories to get the score for Houston to come out to "0.874" and that for NYC to be "0.872". [I think I've done so and get 0.873618 and
0.872729 respectively.]
Add a calculation of the Gini-Simpson diversity index using the five categories of White, Black, Hispanic, Asian, and Other.
Note where the Bay Area stands in terms of the diversity index.
For bonus points:
Deliverable:
msas_df.http://www.census.gov/developers/data/sf1.xml
compare to http://www.census.gov/prod/cen2010/briefs/c2010br-02.pdf
I think the P0050001 might be the key category
P0050002 Not Hispanic or Latino (total) =
Not Hispanic Other (should also be P0050002 - (P0050003 + P0050004 + P0050006)
P0050010 Hispanic or Latino
P0050010 = P0050011...P0050017
From Hispanic and Latino Americans (Wikipedia):
While the two terms are sometimes used interchangeably, Hispanic is a narrower term which mostly refers to persons of Spanish speaking origin or ancestry, while Latino is more frequently used to refer more generally to anyone of Latin American origin or ancestry, including Brazilians.
and
The Census Bureau's 2010 census does provide a definition of the terms Latino or Hispanic and is as follows: “Hispanic or Latino” refers to a person of Cuban, Mexican, Puerto Rican, South or Central American, or other Spanish culture or origin regardless of race. It allows respondents to self-define whether they were Latino or Hispanic and then identify their specific country or place of origin.[52] On its website, the Census Bureau defines "Hispanic" or "Latino" persons as being "persons who trace their origin [to]... Spanish speaking Central and South America countries, and other Spanish cultures".
In the Racial Dot Map: "Whites are coded as blue; African-Americans, green; Asians, red; Hispanics, orange; and all other racial categories are coded as brown."
In this notebook, we will relate the Racial Dot Map 5-category scheme to the P005* variables.
In [1]:
%pylab --no-import-all inline
In [2]:
import numpy as np
import matplotlib.pyplot as plt
from pandas import DataFrame, Series, Index
import pandas as pd
from itertools import islice
In [3]:
import census
import us
import settings
The census documentation has example URLs but needs your API key to work. In this notebook, we'll use the IPython notebook HTML display mechanism to help out.
In [4]:
c = census.Census(key=settings.CENSUS_KEY)
In [5]:
# generators for the various census geographic entities of interest
def states(variables='NAME'):
geo={'for':'state:*'}
states_fips = set([state.fips for state in us.states.STATES])
# need to filter out non-states
for r in c.sf1.get(variables, geo=geo):
if r['state'] in states_fips:
yield r
def counties(variables='NAME'):
"""ask for all the states in one call"""
# tabulate a set of fips codes for the states
states_fips = set([s.fips for s in us.states.STATES])
geo={'for':'county:*',
'in':'state:*'}
for county in c.sf1.get(variables, geo=geo):
# eliminate counties whose states aren't in a state or DC
if county['state'] in states_fips:
yield county
def counties2(variables='NAME'):
"""generator for all counties"""
# since we can get all the counties in one call,
# this function is for demonstrating the use of walking through
# the states to get at the counties
for state in us.states.STATES:
geo={'for':'county:*',
'in':'state:{fips}'.format(fips=state.fips)}
for county in c.sf1.get(variables, geo=geo):
yield county
def tracts(variables='NAME'):
for state in us.states.STATES:
# handy to print out state to monitor progress
# print state.fips, state
counties_in_state={'for':'county:*',
'in':'state:{fips}'.format(fips=state.fips)}
for county in c.sf1.get('NAME', geo=counties_in_state):
# print county['state'], county['NAME']
tracts_in_county = {'for':'tract:*',
'in': 'state:{s_fips} county:{c_fips}'.format(s_fips=state.fips,
c_fips=county['county'])}
for tract in c.sf1.get(variables,geo=tracts_in_county):
yield tract
def msas(variables="NAME"):
for state in us.STATES:
geo = {'for':'metropolitan statistical area/micropolitan statistical area:*',
'in':'state:{state_fips}'.format(state_fips=state.fips)
}
for msa in c.sf1.get(variables, geo=geo):
yield msa
def block_groups(variables='NAME'):
# http://api.census.gov/data/2010/sf1?get=P0010001&for=block+group:*&in=state:02+county:170
# let's use the county generator
for county in counties(variables):
geo = {'for':'block group:*',
'in':'state:{state} county:{county}'.format(state=county['state'],
county=county['county'])
}
for block_group in c.sf1.get(variables, geo):
yield block_group
def blocks(variables='NAME'):
# http://api.census.gov/data/2010/sf1?get=P0010001&for=block:*&in=state:02+county:290+tract:00100
# make use of the tract generator
for tract in tracts(variables):
geo={'for':'block:*',
'in':'state:{state} county:{county} tract:{tract}'.format(state=tract['state'],
county=tract['county'],
tract=tract['tract'])
}
for block in c.sf1.get(variables, geo):
yield block
def csas(variables="NAME"):
# http://api.census.gov/data/2010/sf1?get=P0010001&for=combined+statistical+area:*&in=state:24
for state in us.STATES:
geo = {'for':'combined statistical area:*',
'in':'state:{state_fips}'.format(state_fips=state.fips)
}
for csa in c.sf1.get(variables, geo=geo):
yield csa
def districts(variables="NAME"):
# http://api.census.gov/data/2010/sf1?get=P0010001&for=congressional+district:*&in=state:24
for state in us.STATES:
geo = {'for':'congressional district:*',
'in':'state:{state_fips}'.format(state_fips=state.fips)
}
for district in c.sf1.get(variables, geo=geo):
yield district
def zip_code_tabulation_areas(variables="NAME"):
# http://api.census.gov/data/2010/sf1?get=P0010001&for=zip+code+tabulation+area:*&in=state:02
for state in us.STATES:
geo = {'for':'zip code tabulation area:*',
'in':'state:{state_fips}'.format(state_fips=state.fips)
}
for zip_code_tabulation_area in c.sf1.get(variables, geo=geo):
yield zip_code_tabulation_area
In [6]:
def census_labels(prefix='P005', n0=1, n1=17, field_width=4, include_name=True, join=False):
"""convenience function to generate census labels"""
label_format = "{i:0%dd}" % (field_width)
variables = [prefix + label_format.format(i=i) for i in xrange(n0,n1+1)]
if include_name:
variables = ['NAME'] + variables
if join:
return ",".join(variables)
else:
return variables
def rdot_labels(other=True):
if other:
return ['White', 'Black', 'Asian', 'Hispanic', 'Other']
else:
return ['White', 'Black', 'Asian', 'Hispanic']
FINAL_LABELS = ['NAME', 'Total'] + rdot_labels() + ['p_White', 'p_Black', 'p_Asian', 'p_Hispanic', 'p_Other'] + ['entropy5', 'entropy4', 'entropy_rice', 'gini_simpson']
def convert_to_rdotmap(row):
"""takes the P005 variables and maps to a series with White, Black, Asian, Hispanic, Other
Total"""
return pd.Series({'Total':row['P0050001'],
'White':row['P0050003'],
'Black':row['P0050004'],
'Asian':row['P0050006'],
'Hispanic':row['P0050010'],
'Other': row['P0050005'] + row['P0050007'] + row['P0050008'] + row['P0050009'],
}, index=['Total', 'White', 'Black', 'Hispanic', 'Asian', 'Other'])
def normalize(s):
"""take a Series and divide each item by the sum so that the new series adds up to 1.0"""
total = np.sum(s)
return s.astype('float') / total
def normalize_relabel(s):
"""take a Series and divide each item by the sum so that the new series adds up to 1.0
Also relabel the indices by adding p_ prefix"""
total = np.sum(s)
new_index = list(Series(s.index).apply(lambda x: "p_"+x))
return Series(list(s.astype('float') / total),new_index)
def entropy(series):
"""Normalized Shannon Index"""
# a series in which all the entries are equal should result in normalized entropy of 1.0
# eliminate 0s
series1 = series[series!=0]
# if len(series) < 2 (i.e., 0 or 1) then return 0
if len(series1) > 1:
# calculate the maximum possible entropy for given length of input series
max_s = -np.log(1.0/len(series))
total = float(sum(series1))
p = series1.astype('float')/float(total)
return sum(-p*np.log(p))/max_s
else:
return 0.0
def gini_simpson(s):
# https://en.wikipedia.org/wiki/Diversity_index#Gini.E2.80.93Simpson_index
s1 = normalize(s)
return 1-np.sum(s1*s1)
def entropy_rice(series):
"""hard code how Rice U did calculation
This function takes the entropy5 calculation and removes the contribution from 'Other'
"""
# pass in a Series with
# 'Asian','Black','Hispanic','White','Other'
# http://kinder.rice.edu/uploadedFiles/Urban_Research_Center/Media/Houston%20Region%20Grows%20More%20Ethnically%20Diverse%202-13.pdf
s0 = normalize(series)
s_other = s0['Other']*np.log(s0['Other']) if s0['Other'] > 0 else 0.0
return (np.log(0.2)*entropy(series) - s_other)/np.log(0.25)
def diversity(df):
"""Takes a df with the P005 variables and does entropy calculation"""
# convert populations to int
df[census_labels(include_name=False)] = df[census_labels(include_name=False)].astype('int')
df = pd.concat((df, df.apply(convert_to_rdotmap, axis=1)),axis=1)
df = pd.concat((df,df[rdot_labels()].apply(normalize_relabel,axis=1)), axis=1)
df['entropy5'] = df.apply(lambda x:entropy(x[rdot_labels()]), axis=1)
df['entropy4'] = df.apply(lambda x:entropy(x[rdot_labels(other=False)]), axis=1)
df['entropy_rice'] = df.apply(lambda x:entropy_rice(x[rdot_labels()]), axis=1)
df['gini_simpson'] = df.apply(lambda x:gini_simpson(x[rdot_labels()]), axis=1)
return df
In [7]:
# grab states, convert populations to int
states_df = DataFrame(list(states(census_labels())))
states_df = diversity(states_df)
In [8]:
states_df[FINAL_LABELS].head()
Out[8]:
In [9]:
states_df.sort_index(by='entropy5', ascending=False)[FINAL_LABELS].head()
Out[9]:
In [10]:
r = list(counties(census_labels()))
In [11]:
counties_df = DataFrame(r)
counties_df = diversity(counties_df)
counties_df[FINAL_LABELS].head()
Out[11]:
In [12]:
counties_df.sort_index(by='entropy5', ascending=False)[FINAL_LABELS].head()
Out[12]:
In [13]:
# msas
r = list(msas(census_labels()))
In [14]:
len(r)
Out[14]:
In [15]:
df=DataFrame(r)
df[census_labels(include_name=False)] = df[census_labels(include_name=False)].astype('int')
msas_grouped = df.groupby('metropolitan statistical area/micropolitan statistical area')
#df1 = msas_grouped.apply(lambda x:Series((list(x['NAME']), sum(x['P0050001'])), index=['msas','total_pop'])).sort_index(by='total_pop', ascending=False)
df1 = msas_grouped.apply(lambda x:Series((list(x['NAME']), ),
index=['msas']))
df2 = msas_grouped.sum()
df3 = pd.concat((df1,df2), axis=1)
df3['NAME'] = df3.apply(lambda x: "; ".join(x['msas']), axis=1)
In [16]:
msas_df = diversity(df3)
In [17]:
# grab the ten most populous msas and sort by entropy_rice
msas_df.sort_index(by='Total', ascending=False)[:10].sort_index(by='entropy_rice', ascending=False)[FINAL_LABELS]
Out[17]:
In [18]:
# Testing code
def to_unicode(vals):
return [unicode(v) for v in vals]
def test_msas_df(msas_df):
min_set_of_columns = set(['Asian','Black','Hispanic', 'Other', 'Total', 'White',
'entropy4', 'entropy5', 'entropy_rice', 'gini_simpson','p_Asian', 'p_Black',
'p_Hispanic', 'p_Other','p_White'])
assert min_set_of_columns & set(msas_df.columns) == min_set_of_columns
# https://www.census.gov/geo/maps-data/data/tallies/national_geo_tallies.html
# 366 metro areas
# 576 micropolitan areas
assert len(msas_df) == 942
# total number of people in metro/micro areas
assert msas_df.Total.sum() == 289261315
assert msas_df['White'].sum() == 180912856
assert msas_df['Other'].sum() == 8540181
# list of msas in descendng order by entropy_rice
top_10_metros = msas_df.sort_index(by='Total', ascending=False)[:10]
msa_codes_in_top_10_pop_sorted_by_entropy_rice = list(top_10_metros.sort_index(by='entropy_rice',
ascending=False).index)
assert to_unicode(msa_codes_in_top_10_pop_sorted_by_entropy_rice)== [u'26420', u'35620', u'47900', u'31100', u'19100',
u'33100', u'16980', u'12060', u'37980', u'14460']
top_10_metro = msas_df.sort_index(by='Total', ascending=False)[:10]
list(top_10_metro.sort_index(by='entropy_rice', ascending=False)['entropy5'])
np.testing.assert_allclose(top_10_metro.sort_index(by='entropy_rice', ascending=False)['entropy5'],
[0.79628076626851163, 0.80528601550164602, 0.80809418318973791, 0.7980698349711991,
0.75945930510650161, 0.74913610558765376, 0.73683277781032397, 0.72964862063970914,
0.64082509648457675, 0.55697288400004963])
np.testing.assert_allclose(top_10_metro.sort_index(by='entropy_rice', ascending=False)['entropy_rice'],
[0.87361766576115552,
0.87272877244078051,
0.85931803868749834,
0.85508015237749468,
0.82169723530719896,
0.81953527301129059,
0.80589423784325431,
0.78602596561378812,
0.68611350427640316,
0.56978827050565117])
In [19]:
# you are on the right track if test_msas_df doesn't complain
test_msas_df(msas_df)
In [20]:
# code to save your dataframe to a CSV
# upload the CSV to bCourses
# uncomment to run
# msas_df.to_csv("msas_2010.csv", encoding="UTF-8")
In [21]:
# load back the CSV and test again
# df = DataFrame.from_csv("msas_2010.csv", encoding="UTF-8")
# test_msas_df(df)
In [22]:
all_categories = census_labels('P005',2,10, include_name=False) + \
census_labels('P005',11,17, include_name=False)
all_categories
Out[22]:
In [23]:
msas_df['entropy_all'] = msas_df.apply(lambda x:entropy(x[all_categories]), axis=1)
In [24]:
msas_df.sort_index(by='entropy_all', ascending=False)[FINAL_LABELS + ['entropy_all']][:20]
Out[24]:
In [25]:
msas_df.sort_index(by='P0050001', ascending=False).head()
Out[25]:
In [26]:
top_10_metros = msas_df.sort_index(by='Total', ascending=False)[:10]
top_10_metros['City'] = top_10_metros['NAME'].apply(lambda name: name.split('-')[0])
top_10_metros.sort(columns=['entropy_rice'], inplace=True, ascending=True)
cities = pd.Series(top_10_metros['City'])
diversity = pd.Series(top_10_metros['entropy_rice'])
p_white = pd.Series(top_10_metros['p_White'])
p_asian = pd.Series(top_10_metros['p_Asian'])
p_black = pd.Series(top_10_metros['p_Black'])
p_latino = pd.Series(top_10_metros['p_Hispanic'])
In [27]:
import matplotlib.pyplot as plt
fig = plt.figure(figsize=(10, 8))
ax = plt.subplot(111)
# y axis locations for diversity and races
y_div = np.arange(len(cities))*2
y_race = (np.arange(len(cities))*2)+1
# diversity bars
pDiversity = ax.barh(y_div, diversity, alpha=0.4)
# stacked horizontal bars
pWhite = ax.barh(y_race, p_white, color='b')
pLatino = ax.barh(y_race, p_latino, color='g', left=p_white)
pBlack = ax.barh(y_race, p_black, color='r', left=p_white+p_latino)
pAsian = ax.barh(y_race, p_asian, color='c', left=p_white+p_latino+p_black)
plt.yticks(y_race, cities)
# legend foo https://stackoverflow.com/questions/4700614/how-to-put-the-legend-out-of-the-plot
# Shink current axis's height by 10% on the bottom
box = ax.get_position()
ax.set_position([box.x0, box.y0 + box.height * 0.1,
box.width, box.height * 0.85])
# Put a legend below current axis
ax.legend((pWhite, pLatino, pBlack, pAsian, pDiversity), ('White', 'Latino', 'Black', 'Asian', 'Diversity'),
loc='upper center', bbox_to_anchor=(0.5, -0.05),
fancybox=True, shadow=True, ncol=5)
plt.show()
# If you want to save it