Why his voice matters
Spiegelhalter is, by public standing, the UK statistician most widely recognised as a communicator of statistical reasoning to non-statistician audiences. His professional career has been about building bridges between technical statistical understanding and public reasoning — which is exactly the skill the Letby statistical evidence requires of the CCRC and the Court of Appeal.
When he speaks about statistical evidence in criminal trials, he speaks with the authority of past-president of the Royal Statistical Society, Emeritus Professor at Cambridge, and author of the canonical recent UK popular-reference on statistics. His public commentary on the Letby shift-rota chart is part of the statistical-community consensus against the Crown’s framing.
Professional background
- Emeritus Professor of Statistics, Statistical Laboratory, University of Cambridge.
- Former Winton Professor of the Public Understanding of Risk.
- Past president of the Royal Statistical Society.
- Knighted in 2014 for services to medical statistics.
- Author of The Art of Statistics: Learning from Data (2019), Sex by Numbers (2015), and numerous technical monographs on Bayesian methods.
- Long-standing public-commentator track record on statistical reasoning in medical, criminal, and policy contexts.
The Spiegelhalter statistical-communication frame
Spiegelhalter’s framework for statistical evidence in criminal trials centres on three principles:
- Make the denominator explicit. A pattern is only informative against the denominator it is drawn from. Selection-effect problems arise when the denominator is implicit.
- Distinguish probability of evidence from probability of guilt. The prosecutor’s fallacy confuses P(evidence | innocent) with P(guilty | evidence). This is exactly the error the RSS post-Sally-Clark guidance was designed to prevent.
- Model competing hypotheses, not just one. A statistical analysis that models only the prosecution hypothesis cannot discriminate between it and the alternatives. Bayesian reasoning formally requires the alternatives be modelled.
How the Letby chart fails each principle
The shift-rota chart at the Letby trial fails all three Spiegelhalter principles: the denominator of shifts where Letby was not on was implicit rather than explicit; the probability of shift-overlap given innocence was treated as probability of guilt given shift-overlap; and the natural-cause and systems-failure alternatives were not formally modelled.
This is not a minor critique. It is the same critique Prof. Richard Gill applied to the Lucia de Berk case, that Prof. Peter Green applied via the RSS post-Clark framework, and that the statistical community’s collective response applies to Letby.