def get_vote_outliers(voters):
# Run a really simple statistical model to see which voters don't
# match predicted outcomes.
import numpy
from logistic_regression import logistic_regression, calcprob
# Build a binary matrix of predictors.
predictor_names = ('party', 'ideolog_score')
party_values = { "Democrat": -1, "Republican": 1 }
vote_values = { "+": 1, "-": 0 }
x = [ [] for predictor in predictor_names ]
y = [ ]
for voter in voters:
x[0].append(party_values.get(voter.party, 0)) # independents and unrecognized parties get 0
x[1].append(getattr(voter, 'ideolog_score', 0)) # ideology scores may not be available in a Congress, also not available for vice president
y.append(vote_values.get(voter.option.key, .5)) # present, not voting, etc => .5
x = numpy.array(x)
y = numpy.array(y)
# Perform regression.
try:
regression_beta, J_bar, l = logistic_regression(x, y)
except ValueError:
# Something went wrong. No outliers will be reported.
return
# Predict votes.
estimate = calcprob(regression_beta, x)/100.0
# Mark voters whose vote is far from the prediction.
for i, v in enumerate(voters):
v.is_outlier = (abs(y[i]-estimate[i]) > .7)
def get_vote_outliers(voters):
# Run a really simple statistical model to see which voters don't
# match predicted outcomes.
import numpy
from logistic_regression import logistic_regression, calcprob
# Build a binary matrix of predictors.
predictor_names = ('party', 'ideolog_score')
party_values = { "Democrat": -1, "Republican": 1 }
vote_values = { "+": 1, "-": 0 }
x = [ [] for predictor in predictor_names ]
y = [ ]
for voter in voters:
x[0].append(party_values.get(voter.party, 0)) # independents and unrecognized parties get 0
x[1].append(getattr(voter, 'ideolog_score', 0)) # ideology scores may not be available in a Congress, also not available for vice president
y.append(vote_values.get(voter.option.key, .5)) # present, not voting, etc => .5
x = numpy.array(x)
y = numpy.array(y)
# Perform regression.
try:
regression_beta, J_bar, l = logistic_regression(x, y)
except ValueError:
# Something went wrong. No outliers will be reported.
return
# Predict votes.
estimate = calcprob(regression_beta, x)/100.0
# Mark voters whose vote is far from the prediction.
for i, v in enumerate(voters):
v.is_outlier = (abs(y[i]-estimate[i]) > .7)