Daw's miraculous statistics are worthless

Two small igneous erratics found near the higher path on Baggy Point.  Altitudes c 45m and c 60m. These were in the view of Daw too high to be counted as low, and too low to be counted as high.  They were "rejected" becausee they have been moved at least once, and because they do not have "a secure geological context."   But does any free glacial erratic have a secure context?  They are called "erratics" for a perfectly good reason, and almost all of those on farmland will have been moved by farmers at some stage..........  (Photos:  courtesy Paul Madgett)

In the context of ongoing attacks on my integrity and competence, ex-Stonehenge tour guide Tim Daw likes to portray himself as a feisty defender of the truth. But in the welter of his recent publications dealing with the Devon coast he has become increasingly dependent upon AI bots. He has also corrupted statistical methods to make inconvenient erratics disappear and has invented more than 30 “phantom erratics” (all, purely by chance, at 5m OD) so as to boost his shoreline data set and to achieve his desired results.

It’s amazing what lengths some people will go to in order to create pseudo-scientific “proofs” for their rather wacky hypotheses. And it's also amazing — and more than a little sad — that they often get away with it…….

In the recent article by Daw, Ixer and Madgett much emphasis is placed on the supposed existence of just one “high level erratic” in the Baggy Point - Barnstaple area, referred to as the Ramson Cliff erratic. It’s seen as an outlier or an anomaly, with all other documented erratics lying on or around the altitude of the shore platform. That is a false representation of the situation, since there are abundant recorded erratics between 30m and 80m showing that glacier ice affected much of the local coastal environment. Undeterred, Daw asked Grok for “an independent analysis” of his own paper, and for a “Bayesian analysis” of the likelihood of human transport over glacial transport.” As ever with AI bots, Grok faithfully delivered what was expected by its master……...

Some other examples from Daw, who is currently on a mission to demonstrate to the world that active glacial ice cannot possibly have had anything to do with the erratics scattered along the coastal zone of Devon and Cornwall.

The Myth of Bristol Channel High-level Glacial Erratics (December 2024): Considers the lack of a "glacial imprint" on boulders and argues their placement is more likely due to residual weathering than glacial action. This is not a statistical study, but in supposedly analysing a list of “cited erratics” Daw simply decides that each of them is irrelevant because they are too low, or of the wrong rock type, or “non-glacial”, or lacking in cited evidence, or lacking in geological context. That’s absurd. The process is entirely subjective and lacking in any appreciation of glacial environments or processes.

Analysis of Claims Regarding High-Level Glacial Erratics (April 2025): Uses a quantitative list of erratics and their altitudes to disqualify them as evidence of an ice-sheet override. This is essentially the same article, but here published on Researchgate without any peer review. See the above comments.

ChatGPT on competing transport hypotheses (August 2025) Monthly Archive (specifically the AIC section): Includes a section where ChatGPT was reportedly used to put competing transport hypotheses into a statistical model comparison (AIC) to show that a non-glacial model better explains the body of evidence. This is a nonsensical exercise, based on biased data sets and observations fed into the AI model. Rubbish in, rubbish out.

The Statistical Improbability of a Ramson Cliff Glacial Erratic (October 2025): This post uses a table of "high-level" erratics to argue that their scarcity at specific altitudes makes a glacial origin statistically unlikely. The data set is completely unacceptable — Daw seems to have chosen some erratics from the literature and to have ignored others that were deemed inconvenient. He added several inland erratics from the Fremington area — all at 25m — while completely ignoring the records of other erratics in the sediments near Eastacombe up to 90m. Then he added “approximately 30 other foreshore erratics at 5m OD ”for statistical purposes” in order to bulk up his shore platform group! So he ended up with a synthetic data set. Malpractice. As far as we know, these additional erratics might not even exist. His graph of the elevations of 14 chosen erratics is also nonsensical.

Caution in Attributing the Fremington Clay Series (October 2025): Daw argues that the abundance of local Devon and Cornish erratics over far-travelled ones is "statistically consistent" with a local origin rather than a distant glacial source. Revised and expanded on 4 April 2026.   Daw gets into a frightful tangle over the altitude of the ice dam that held up a lake in the Fremington area, and seems to assume, bizarrely, that the surface of the lake was at the same altitude as the varved lake deposits that are known in the literature, up to 35m OD. That implies a lake surface level at 65m, or even higher. He goes out of his way to avoid referring to any of the Fremington deposits as glacially deposited, and misrepresents the views of those who have described the Fremington till in detail. He mentions fluvioglacial deposits at c 55m OD and still insists that the Ramson Cliff erratic, at c 85m OD had nothing to do with glaciation.

In these articles, and others, Tim Daw seeks to use statistics to “prove” that the coastal erratics around Croyde and Saunton are ice-rafted, and that the Ramson Cliff erratic, at about 85 m above sea level, is an outlier that cannot possibly have been transported by ice. It all looks terribly impressive and scientific, until you realise that his methods are the stuff of statistical nightmares……...

I’m no statistician, but I’m not stupid, and it is obvious from the outset that Daw's methods are corrupt. What we see is a classic case of selection bias (called “data scrubbing” in the trade) to achieve a predetermined result.

For example, in his convoluted attempts to prove that the Ramson Cliff boulder is an outlier, these are the statistical red flags:

1. Sampling Bias (Cherry-Picking)
Daw chooses to ignore or exclude the 45m and 60m erratics near Baggy Point (recorded by Berry and Madgett) on the grounds that they may have been moved and that they do not have “a secure geological context”. That immediately introduces subjectivity and selection bias. You could argue that every erratic in a farmed landscape has probably been moved at some stage, and that every isolated boulder on the shore platform has also been moved by the waves — and that they should all be eliminated from the study on that basis………) Daw also chooses to ignore the 90m till at Eastacombe, although it is recorded by the BGS and is known from borehole records to contain erratics. So Daw is deliberately "truncating the distribution” by eliminating all boulders that he deems to refer to as “contested” and including all those that he deems to be “uncontested”. That is, to put it mildly, completely unprofessional. Further, if you consciously remove all data points between 25m and 75m, the 80m point will mathematically appear to be an extreme outlier. This is not a discovery; it is a result of the filtering or sampling process itself.

2. Circular Reasoning (Begging the Question)
Daw labels all high-level stones as "contested" or "anthropogenic" because they don't fit the low-level ice rafting model, and then uses that "cleaned" model to prove high-level glacial erratics don't exist. This is a "Self-Fulfilling Prophecy." You cannot use a model to validate the exclusion of the very data that would disprove the model.

3. Arbitrary Binning
Creating a “shoreline group” (clustered around 5–25m) and a “high level group" based on some arbitrary altitude, rather than assuming the presence of a continuous gradient, is known to statisticians as “arbitrary binning."  This creates a false dichotomy. In nature, glacial deposition follows strict principles (not yet fully understood) but apparently random within a range. By forcing the data into two bins and ignoring the middle, Daw is guilty of manufacturing a "gap" that does not exist in the raw geological or geomorphological record.

4. Small Sample Size & Power
By using a small pool of "verified” or “uncontested” surface boulders of certain dimensions and shapes, while ignoring the presence of inconvenient boulders and cobbles found in exposures and sub-surface boreholes at Fremington and elsewhere, Daw reduces the “statistical power" of the study. A study that ignores the most robust institutional data (BGS) in favor of a curated or self-selected list of surface stones is prone to a fundamental error—falsely identifying a natural or expected occurrence as an anomaly or outlier.

5. Confirmation Bias in Bayesian Analysis
Daw uses Bayesian Inference to claim a high probability of human transport in the case of the Ramson Cliff erratic. Bayesian outcomes are highly sensitive to initial assumptions. If your starting assumption is that ice never reached 80m in the Baggy Point area, the mathematics will always tell you an 80m stone is an anomaly. “Rubbish in, rubbish out…………." 

Terribly impressive, until you realise that it's all manipulated nonsense.......

================

In short, Daw is misrepresenting his selection criteria, and pretending that he is telling us something important about the natural population of erratics in the Baggy Point - Barnstaple area. It all looks terribly impressive and very scientific, but it is essentially garbage.