338Canada
Philippe J. Fournier publishes pollster ratings page1. Fournier's public methodology page describes a mostly proportional swing model with regional adjustments.
Philippe J. Fournier publishes pollster ratings page
PASS 4 — The Researchers, The Natural Experiments, and the T
Fournier's public methodology page describes a mostly proportional swing model with regional adjustments
338Canada is a single dominant aggregator
PASS 4 — The Researchers, The Natural Experiments, and the T
TheWrit.ca/338Canada are the closest to a Canadian equivalent of FiveThirtyEight
PASS 4 — The Researchers, The Natural Experiments, and the T
there is no methodological consensus in Canadian psephology on whether to correct for cumulative known-directional bias
Pass 2: The Forensics, the Math, and the Philosophy of Canad
the model weights polls by recent accuracy
Pass 2: The Forensics, the Math, and the Philosophy of Canad
Canadian by-elections expose systematic riding-level model error that is invisible at the national-aggregate level
PASS 4 — The Researchers, The Natural Experiments, and the T
Toronto–St. Paul's was an uninstrumented event
PASS 4 — The Researchers, The Natural Experiments, and the T
the binomial standard error is √(0.40·0.60/1500) = 1.26 percentage points
Pass 2: The Forensics, the Math, and the Philosophy of Canad
These are things people say about 338Canada that aren't settled fact — each one attributed to who's actually claiming it, so a claim never quietly passes as established just because it showed up in a document. Weigh it yourself.
"the model is mechanically herd-following"
"338Canada's outlier-penalty algorithm should be publicly audited"