Added Webcast 4 materials

webcast4/Untitled0.ipynb

@@ -0,0 +1,7 @@
+{
+ "metadata": {
+  "name": "Untitled0"
+ }, 
+ "nbformat": 2, 
+ "worksheets": []
+}

webcast4/multimode.py

@@ -0,0 +1,44 @@
+import networkx as net
+import matplotlib.pyplot as plot
+from collections import defaultdict
+
+def plot_multimode(m,layout=net.spring_layout, type_string='type', with_labels=True, filename_prefix='',output_type='pdf'):
+
+    ## create a default color order and an empty color-map
+    colors=['r','g','b','c','m','y','k']
+    colormap={}
+    d=net.degree(m)  #we use degree for sizing nodes
+    pos=layout(m)  #compute layout 
+
+    #Now we need to find groups of nodes that need to be colored differently
+    nodesets=defaultdict(list)
+    for n in m.nodes():
+        try:
+            t=m.node[n][type_string]
+        except KeyError:
+            ##this happens if a node doesn't have a type_string -- give it a None value
+            t='None'
+        nodesets[t].append(n)
+
+    ## Draw each group of nodes separately, using its own color settings
+    print "drawing nodes..."
+    i=0
+    for key in nodesets.keys():
+        #ns=[d[n]*100 for n in nodesets[key]]
+        net.draw_networkx_nodes(m,pos,nodelist=nodesets[key], node_color=colors[i], alpha=0.6) #node_size=ns,
+        colormap[key]=colors[i]
+        i+=1
+        if i==len(colors): 
+            i=0  ### wrap around the colormap if we run out of colors
+    print colormap  
+
+    ## Draw edges using a default drawing mechanism
+    print "drawing edges..."
+    net.draw_networkx_edges(m,pos,width=0.5,alpha=0.5)  
+
+    print "drawing labels..."
+    if with_labels: 
+        net.draw_networkx_labels(m,pos,font_size=12)
+    plot.axis('off')
+    if filename_prefix is not '':
+        plot.savefig(filename_prefix+'.'+output_type)

webcast4/pac_types.py

@@ -0,0 +1,18 @@
+pac_types_str="""
+C = Communication Cost
+D = Delegate
+E = Electioneering Communication
+H = House
+I = Independent Expenditor (Person or Group)
+N = PAC - Nonqualified
+O = Independent Expenditure-Only (Super PACs)
+P = Presidential
+Q = PAC - Qualified
+S = Senate
+U = Single Candidate Independent Expenditure
+X = Party Nonqualified
+Y = Party Qualified
+Z = National Party Nonfederal Account 
+""".replace(' = ','=').strip().split('\n')
+
+pac_types=dict([tuple(row.split('=')) for row in pac_types_str])

webcast4/two_mode.py

@@ -0,0 +1,73 @@
+#!/usr/bin/env python
+# encoding: utf-8
+"""
+two_mode.py
+
+Created by Maksim Tsvetovat on 2011-08-17.
+Copyright (c) 2011 __MyCompanyName__. All rights reserved.
+"""
+
+import sys
+import os
+import csv 
+import math
+import networkx as net
+import matplotlib.pyplot as plot
+
+## Import bi-partite (bi-modal) functions
+from networkx.algorithms import bipartite as bi
+
+def trim_edges(g, weight=1):
+	g2=net.Graph()
+	for f, to, edata in g.edges(data=True):
+		if edata['weight'] > weight:
+			g2.add_edge(f,to,edata)
+	return g2
+
+
+## Read the data from a CSV file
+## We use the Universal new-line mode since many CSV files are created with Excel
+r=csv.reader(open('campaign_short.csv','rU'))
+
+## 2-mode graphs are usually directed. Here, their direction implies money flow
+g=net.Graph()
+
+## we need to keep track separately of nodes of all types
+pacs=[]
+candidates=[]
+
+## Construct a directed graph from edges in the CSV file
+for row in r: 
+    if row[0] not in pacs: 
+        pacs.append(row[0])
+    if row[12] not in candidates: 
+        candidates.append(row[12])
+    g.add_edge(row[0],row[12], weight=int(row[10]))
+
+## compute the projected graph
+pacnet=bi.weighted_projected_graph(g, pacs, ratio=False)
+pacnet=net.connected_component_subgraphs(pacnet)[0]
+weights=[math.log(edata['weight']) for f,t,edata in pacnet.edges(data=True)]
+
+net.draw_networkx(p,width=weights, edge_color=weights)
+
+
+
+## Compute the candidate network
+cannet=bi.weighted_projected_graph(g, candidates, ratio=False)
+cannet=net.connected_component_subgraphs(cannet)[0]
+weights=[math.log(edata['weight']) for f,t,edata in cannet.edges(data=True)]
+plot.figure(2) ## switch to a fresh canvas
+net.draw_networkx(cannet,width=weights, edge_color=weights)
+
+
+plot.figure(3)
+plot.hist(weights)
+
+## The weights histogram is logarithmic; we should compute the original weight = e^log_weight
+cannet_trim=trim_edges(cannet, weight=math.exp(0.9))
+
+plot.figure(4)
+## re-calculate weights based on the new graph
+weights=[edata['weight'] for f,t,edata in cannet_trim.edges(data=True)]
+net.draw_networkx(cannet_trim,width=weights, edge_color=weights)

webcast4/webcast.py

@@ -0,0 +1,133 @@
+import networkx as net
+from networkx.algorithms import bipartite as bi
+
+from collections import Counter
+
+candidates={}
+can_file=open('2012/foiacn.txt','rb')
+for line in can_file:
+    cid=line[0:9]
+    name=line[9:47].strip()
+    party=line[47:50]
+    inc=line[56]
+    zip=line[147:152]
+    candidates[cid]={'name':name,'party':party,'type':ctype,'inc':inc,'zip':zip}
+
+[(k,pacs[k]) for k in pacs.keys()[:10]]
+
+ctype_counter=Counter()
+for can in candidates.values():
+    ctype_counter[can['type']]+=1
+ctype_counter
+
+from pac_types import pac_types
+pacs={}
+pac_file=open('2012/foiacm.txt','rb')
+for line in pac_file:
+    pid=line[0:9]
+    ctype=line[0]
+    name=line[9:99].strip()
+    party=line[232-235]
+    ctype=line[231]
+    zip=line[225:230]
+    pacs[pid]={'name':name,'party':party,'type':ctype,'zip':zip}
+
+ctype_counter=Counter()
+for pac in pacs.values():
+    ctype_counter[pac['type']]+=1
+ctype_counter
+
+
+g=net.Graph()
+can_list=[]
+pac_list=[]
+contrib=open('2012/itpas2.txt','rb')
+for line in contrib:
+    pid=line[0:9]
+    cid=line[52:61]
+    #amt=int(line[36:43])
+    g.add_edge(pid,cid)
+    if cid not in can_list: can_list.append(cid)
+    if pid not in pac_list: pac_list.append(pid)
+    if pid in pacs: 
+        g.node[pid]=pacs[pid]
+    else:
+        pacs[pid]={'type':'unknown'}
+    if cid in candidates: 
+        g.node[cid]=candidates[cid]
+    else:
+        candidates[cid]={'type':'unknown'}
+
+
+cannet=bi.weighted_projected_graph(g, can_list, ratio=False)
+
+def trim_edges(g, weight=1):
+	g2=net.Graph()
+	for f, to, edata in g.edges(data=True):
+		if edata['weight'] > weight:
+			g2.add_edge(f,to,edata)
+			g2.node[f]=g.node[f]
+			g2.node[to]=g.node[to]
+	return g2
+
+import multimode as mm
+
+cancore=trim_edges(cannet, weight=50)
+mm.plot_multimode(cancore, type_string='party')
+
+pacnet=bi.weighted_projected_graph(g, pac_list, ratio=False)
+paccore = trim_edges(pacnet, weight=50)
+
+
+def sorted_map(dct):
+    ds = sorted(dct.iteritems(), key=lambda (k,v): (-v,k))
+    return ds
+
+d=sorted_map(net.degree(paccore))
+c=sorted_map(net.closeness_centrality(paccore))
+inf_pacs=[pacs[pid] for pid,deg in d[:10]]
+close_pacs=[pacs[pid] for pid,deg in c[:10]]
+
+
+"""
+[{'name': 'NATIONAL ASSOCIATION OF REALTORS POLITICAL ACTION COMMITTEE',
+  'party': ' ',
+  'type': 'Q',
+  'zip': '60611'},
+ {'name': 'AT&T INC. FEDERAL POLITICAL ACTION COMMITTEE (AT&T FEDERAL PAC)',
+  'party': ' ',
+  'type': 'Q',
+  'zip': '75202'},
+ {'name': 'UNITED PARCEL SERVICE INC. PAC',
+  'party': ' ',
+  'type': 'Q',
+  'zip': '30328'},
+ {'name': 'HONEYWELL INTERNATIONAL POLITICAL ACTION COMMITTEE',
+  'party': ' ',
+  'type': 'Q',
+  'zip': '20001'},
+ {'name': "LOCKHEED MARTIN CORPORATION EMPLOYEES' POLITICAL ACTION COMMITTEE",
+  'party': ' ',
+  'type': 'Q',
+  'zip': '22202'},
+ {'name': 'NATIONAL BEER WHOLESALERS ASSOCIATION POLITICAL ACTION COMMITTEE',
+  'party': ' ',
+  'type': 'Q',
+  'zip': '22314'},
+ {'name': 'GENERAL ELECTRIC COMPANY POLITICAL ACTION COMMITTEE (GEPAC)',
+  'party': ' ',
+  'type': 'Q',
+  'zip': '20004'},
+ {'name': 'COMCAST CORPORATION POLITICAL ACTION COMMITTEE- FEDERAL',
+  'party': ' ',
+  'type': 'Q',
+  'zip': '19103'},
+ {'name': 'THE BOEING COMPANY POLITICAL ACTION COMMITTEE',
+  'party': ' ',
+  'type': 'Q',
+  'zip': '22209'},
+ {'name': 'VERIZON COMMUNICATIONS INC./VERIZON WIRELESS GOOD GOVERNMENT CLUB (VERIZON/VERIZON WIRELES',
+  'party': ' ',
+  'type': 'Q',
+  'zip': '20005'}]
+  """