IvanV wrote: ↑Tue Oct 03, 2023 3:33 pm
snoozeofreason wrote: ↑Tue Oct 03, 2023 1:29 pm
I supect that the CPS will argue that none of the evidence in the case was "Statistical" because no numbers were presented to the jury and no statistician gave evidence.
Instead the jury were simply invited to make their own judgements as to whether similarities and patterns observed in the events that unfolded were suspicious or just coincidental. To my mind those are statistical arguments, even if no numbers or statistical terms are used. The jury were effectively left to rely on their intuition. That's exactly what David Spiegelhalter (author of Ivan's quote) was warning against.
I wondered if that is why the professor is criticising the defence case also, for failing to bring out the point that intuition is very faulty in this situation.
In fact Professor Gill has
blogged and he seems to be saying just that, among other things. He seems to be implying that the prosecution deliberately avoided the statistical evidence, once apprised of it, because it was not helpful to their case.
The RSS report is available here. And by implication, the defence could have done better to bring it up. But he also thinks the defence was underfunded to be able to put up a proper defence. He has also been researching a lot of things about the case outside his area of expertise, though he admits that.
So I've had a look through this material stuff now. The RSS report only sets out principles and looks at past cases, it does not analyse the statistics in the Letby case, disappointingly. But I think they are clearly saying someone should have done it.
Gill, one of the statistics professors behind the RSS report, does not present any statistical analysis in his blog, he only points to the RSS report he contributed to. He reproduces strong criticisms of the medical evidence presented, which he brings from an external anonymous source, links available there. But we do have some statistics we can look at in links Gill provides, from one analyst. Gill points to
link and
another link. A bit disappointing that Gill said nothing about it beyond point to it. Maybe given his RSS involvement he feels constrained to do no more than that.
Obviously statistics like this can rarely if ever prove someone innocent. But they can show that there is nothing unusual here that it has to be something other than chance. And, it would seem, that is just what the statistics do say. And the statistics in these links are quite simple, an A-level student could reproduce them. Maybe they are too simple. We learn:
- The 7 deaths which are argued to be the Letby "cluster" would occur with probability 1/83, if we model them as a Poisson distribution, which seems suitable. So just the kind of thing that will occur quite often by chance from time to time and from place to place, given that time passes and there are many hospitals.
- The neonatal death rate at the hospital actually went up in the period immediately following Letby's removal from the ward.
- When eventually the death rate went down, that coincided with the hospital reducing its facilities and no longer taking on such risky cases.
In other words, from a statistical perspective, there really was nothing there you could identify as so unusual it had to be something other than chance. Talk of an unusual clusters of deaths here is, statistically, quite unwarranted, if we believe this analysts stats.