Wednesday, January 16, 2013

Poll aggregation

  1. Poll aggregation based on a Bayesian fixed house-effects model, over a six month span, with 100,000 MCMC iterations. House effects across the six polling houses are constrained to sum to zero
  2. Poll dates are set to the median date of the polling period
  3. House outcome estimates based on a national probability model

No comments:

Post a Comment