Professor Fenton & LBC screw-tubed for breaching community lyinguide

Hi folks, @Everyman posted this a couple of weeks ago:

the uchewb embedded has now been blanked for truth telling which of course breaches the tick box feckers’ view of their lying guide.

I suggest we all seek to avoid that toxic platform if we can. Here’s the latest from the new Norman with Del Bigtree:

@Evvy_dense are these stats the missing evidence in your discussions with PP or is this something totally different?

cheers

1 Like

Just to reinforce what you said @CJ1

2 Likes

Hi CJ. Great find, and thanks for the tag. I’ll just put up @Everyman 's Fenton interview video with Majid Nawaz on LBC, which was deleted (‘screw-tubed’, lol) from Uchube.
Bitchute to the rescue:

I looked at the start of this and it provides useful context so I’ll start there.

There is a lot of detail covered in the Highwire video, though the content of the two videos is similar.

I realize not everyone will want to watch the much longer Highwire video but I hope I can lay out what the key substance of the matter is, and where the detail is to be found.

Near the beginning, in the bitchute link…

Prof. Norman Fenton: "The only true way to evaluate the risk benefit ratio is by comparing the all cause mortality in the vaccinated and unvaccinated. If the virus is as dangerous as claimed, and the vaccine is as effective as claimed, then we should have enough evidence now that it is saving more people’s lives than it is killing.

So the ONS have been publishing data on the England mortality of the vaccinated against unvaccinated. We’ve done detailed analyses of their most recent report".

Comparing all-cause mortality data avoids problems with definition of illnesses, viruses, cases etc. People either died or they didn’t. As we will see, that only leaves one little problem - defining who was vaccinated and who was unvaccinated.

It soon pops up in the bitchute video, that the lead author of this report, Martin Neil, has this pinned tweet:

“Our research team have now analysed the ONS England November mortality data. We conclude that despite seeming evidence to support vaccine effectiveness this conclusion is doubtful because of a range of serious inconsistencies and anomalies.”

This tweet includes a picture of the graph in Figure 16 of the paper

“Latest statistics on England mortality data suggest systematic mis-categorisation…”

ie the main paper discussed in the Highwire video below.

Let’s call this Fenton’s Figure 16, although the lead author is Neil.

What’s kicking about that started this off, is government/ONS graphs ‘showing’ unvaccinated deaths higher than vaccinated deaths, measured by all-cause mortality. It certainly looked good for the pro-vax camp:

image

But the above graph from Fenton’s team (their Figure 16 mentioned above) turns this vaccination “happy ending” attempt by ONS on its head.

Further, it is obtained after careful analysis of the ONS data revealed some anomalies, and the adjusting for those anomalies.

It’s a moment of high drama in the covid-19 vaxed v unvaxed graph arena - for the opposing teams’ graphs can not both be correct. Even an honourable draw looks unlikely. One team is wrong and the outcome of the vaccine risk benefit seems to depend on it.

The Fenton’s Figure 16 graph (one of three, for the age groups 60-69, 70-79 and 80+) indicates a spike in deaths shortly after vaccination, which is not revealed in government data - which counts people as unvaccinated until 14 days after their first dose (and similarly, counts as first dose until 14 days following the second dose has elapsed).

Such a mis-categorisation obviously has a double effect, because what would be a spike in vaccination deaths is allocated by ONS to the unvaccinated - like a goal scored being awarded to the team that conceded it.

It is particularly significant because the first 14 days appears to be the period where post-vaccination deaths are particularly likely.

The mis-categorisation, if this is correct, thereby performs multiple functions in the service of concealing vaccine damage. A critical spike goes to the opposite ‘team’, and is thereby concealed - and of course in the resulting data, the unvaccinated not only take the numerical hit, but bear the political flak, and the resulting costs.

ED

2 Likes

CJ: “@Evvy_dense are these stats the missing evidence in your discussions with PP or is this something totally different?”

Possibly. PP argued that if there were significant deaths due to the vaccine then they would show up in mortality data; that’s why he was keen on that data so he agreed with Fenton’s group about that being the place to look, as did I (I hope). Though I also think the medical evidence is important too, as mortality data took months to appear.

Also, even the possible scenario where vaccines reduced covid deaths but caused other deaths to a similar extent is highly problematic due to the existence of early antiviral treatments that would also prevent covid deaths but much more safely.

But yes, if the adjustments Fenton and team want to make to the ONS all-cause mortality data are correct, it shows where significant vaccine deaths could be hiding (make that, ‘being hidden’ :wink: :astonished:)

When the vaccine rollouts started country by country, it was usually in the elderly, and often in care homes. These early roll-outs did seem to be accompanied by reports of outbreaks of deaths, which because it was care homes, emerged through local reports.
Of course once the vaccinations went down the ages, any outbreaks would not be so obvious as they wouldn’t be localised

2 Likes

EDITING

Some detail on CJ’s link which started the thread (I think I’ve given some of the gist above).
A sort of transcript.
Finish tomorrow hopefully.
ED.

Hi @Evvy_dense thanks for looking at this in detail - if Fenton et al are correct then the house of Vax would seem to be in great danger of collapsing.

I suppose the ONS could claim they’re right because the figures show the government’s policy of frightening the unvaxed to death is working! :wink:

cheers

[Minor edits added] Prof Norman Fenton on the Highwire with Del Bigtree. (CJ’s link)

As I say I’ve a transcript of sorts, which I’ll post underneath.

Some comments about the highlights, as they don’t beat a very direct path to the main goods (and they ARE main), and there is other stuff going on.

The crux of their argument is:

*Vaccination deaths occurring within 2 weeks of a jab appear to be being buried by counting them as unvaccinated deaths, due to the construction of the data solely to reflect vaccine efficacy, which needs a two week waiting period before the vaccine is effective.
*The vaccination roll-out appears to be associated with peaks in all-cause mortality in the unvaccinated, which is obviously strange.
*The vaccination roll-out was staggered by age - yet in all three age groups the age-dependent peaks coincided with the peak in the ‘unvaccinated’ deaths.
*Once you postulate this mis-categorization it’s natural to adjust for it; this seems to show a spike in vaccine deaths after the first dose

There are a lot of graphs and tables in the video; I’ve only copied the necessary ones into the transcript. The narrative below is a list of the tables/graphs in the transcript - for reference. This might help keep track of what’s what, and where you are (or it might not :slightly_frowning_face:)

“Table 2”. Table of covid-19 by vaccination rates. By age group.

All fine, except that in groups over 30 years old the rate among the vaccinated is much higher than the unvaccinated, about double on average.

A quick look at a graph of the 10-59 age group {a truly r-i-d-i-c-u-l-o-u-s age group for PHE to issue} which shows all-cause mortality significantly higher in the vaccinated - BUT which is misleading as most of those dying in this age group are near the top of it anyway, and so most would be vaccinated.
(It’s a good warning - always consider whether the two groups compared in one respect are genuinely similar in others.)

Then there is a trio of all-cause mortality graphs 60-69, 70-79 and 80+. These appear to show unvaccinated deaths higher than vaccinated deaths!

However each shows a strange peak in unvaccinated deaths around the vaccine rollout.

Then there is a graph of non-covid deaths, which still shows a peak in the unvaccinated around vaccination; this graph is there to answer the suggestion that covid was somehow causing those peaks in the unvaccinated deaths - Fenton explains that that it wasn’t.

They then look at a couple of artificial examples (in two tables) in order to make a specific and rather startling point - separate from the point about mis-categorization - showing how another seemingly ‘minor’ inaccuracy in the data (a week long delay in the death reporting) can produce wholly misleading or wrong results, which actually resemble ONS’s vaccine-friendly results.

They thereby show in a graph ironically titled ‘Placebo apparently saves lives’ (that is similar to actual ONS graphs), of this MADE-UP placebo data, how a vaccine that is placebo can appear to be a big lifesaver, and yet only require a small delay in death reporting.

Along with the miscategorization, this is a second way to produce spectacular-seeming reduction in deaths from a vaccine that is the same as a placebo.

The next graph “Mathematician uncovers disturbing mortality data” is of the mortality in the different categories of vaccination.

It’s non-covid deaths, which means that all three graphs should be the same. And they have that data from Life tables, insurance industry presumably, to check against.

But they aren’t the same - they are very different. One of the three has mortality at a fraction of expected. It is the double jabbed - which apparently makes you ‘immortal’, Fenton observes ironically.

There is then another digression for another anomaly. Why do they DO that? (Fenton). Seems he’s got a million of 'em.
It’s a table; might be worth skipping if you’re getting fed up - best to stay awake for the last item.

Lastly, the piece de resistance which they didn’t actually trumpet much, and which isn’t well explained - as the explanation is disrupted by some cross-chat. It confused me, but fortunately I could refer to the paper, so here’s the gist of what they did:

Having established, or at least strongly posited, the mis-categorization (“vaxed, died within 2 weeks, that’ll be unvaccinated, then”) and the anomaly caused by it, any modellers worth their salt will then want to see if they can adjust for it in order to obtain correct, or at least more reasonable, data.

So using expected deaths in the unvaccinated (from the Life tables) they take the seemingly erroneous extra deaths (which can’t really be due to the vaccine) and re-allocate them where they probably should have been in the first place - with the vaccinated.

This indicates that in reality, among the vaccinated, there is a peak in excess mortality at the first dose.
That’s the last graph, the fabled (well I fabled it in the introductary post) Figure 16 of the paper where the spike can be seen.

After discussion, Del Bigtree puts up the conclusions (which seemed worded differently from their papers) and then they round it up with some chat.

All a bit of a ‘phew’.

Although Fenton is well placed by his qualifications and experience to do this kind of analysis, it has come under attack on the new top-reach journal (Twitter). But according to Fenton and his team the analysis hasn’t been subjected to any serious academic questioning. Personal attacks though, yeah
It does look like a complete analysis, and it is undeniable that the ONS data look suspiciously contrived or surprisingly inept.

I’m really posting the rough transcript now :slight_smile:

ED

1 Like

Prof Norman Fenton on the Highwire with Del Bigtree - Transcript

The main paper discussed is:

Martin Neil at al, Latest statistics on England mortality data suggest systematic mis-categorisation of vaccine status and uncertain effectiveness of Covid-19 vaccination. (December 2021)

https://www.researchgate.net/publication/356756711_Latest_statistics_on_England_mortality_data_suggest_systematic_mis-categorisation_of_vaccine_status_and_uncertain_effectiveness_of_Covid-19_vaccination

Some comments in curly brackets. Yes, you can’t shut some people up :slightly_smiling_face: (but most are meant to be navigational).

ED

(About 3m). Shows 3 other papers from this team, which seems to be headed by Prof Fenton. They are linked at the bottom of above paper. Generally the papers ask questions of the narrative itself, or at least the numbers.

Jumping in a few mins after the start, this is a sort of transcript;

Prof Fenton: Around Sep 2020. The ramping up of testing of asymptomatic people, puzzled us.

(6m in) We noticed published data didn’t match with acquired data from the ground. Papers querying this, couldn’t get them published - were rejected without review. Couldn’t even get it accepted onto preprint servers, which accept anything, that’s the whole point!

We put it on ResearchGate, it’s actually very well read there. That’s good enough, we just want it to be read. It culminated in this vaccination stuff. We weren’t looking at vaccinations, and didn’t have any strong view.

Maclachlan. Well-known VAERS paper.

https://www.researchgate.net/publication/352837543_Analysis_of_COVID-19_vaccine_death_reports_from_the_Vaccine_Adverse_Events_Reporting_System_VAERS_Database_Interim_Results_and_Analysis

{This paper, also from Fenton’s team, raised the profile of VAERS reports, still dismissed as ‘anecdotal’}

We were suspicious of these (vaccine) studies, the randomized trials and also the observational studies claiming the vaccine was very effective at stopping transmission. We saw all kinds of problems with these studies and we felt the best way … was to look at all-cause mortality over a period of time. We should clearly be seeing fewer deaths among the vaccinated than the unvaxed.

This completely bypasses all the problems with the other studies, where you worry about definitions of case, etc.

Adverse reactions after vaccn often happen quickly so it’s wrong to count as unvaxed after vaccination.

First look at the data as presented:
image

Table 2 from here, published by PHE (maybe HSA now) https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/1027511/Vaccine-surveillance-report-week-42.pdf {Data up to 17 October 2021}

{Highlights the last 2 cols}. Later claims of ‘wrong data, wrong population’ etc, and vaccine stops you from dying, or from being hospitalized - so, yeah, you can look at mortality data.

The ONS has published the report that we focus on, that we did our analysis on. Published last month, the data on mortality

https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/bulletins/deathsinvolvingcovid19byvaccinationstatusengland/deathsoccurringbetween2januaryand24september2021

Age categories were crude:

10-59
60-69
70-79
80+

They didn’t provide a proper breakdown of 10-59 age category.

DB: (Ironic) Let’s go 10-59? And hardly anyone was dying in <18, it’s hard to imagine why you would lump them in with people aged 59.

NF And remember they only provided this because we put in a FoI request, and also spoke to the authors of the previous version of this report.

But at least we’ve got something decent we can work on for the older age categories.

Let’s {briefly} look at the 10-59 age category.

This is a plot each week of the number of deaths per 100,000 comparing the vaccinated to the unvaccinated.

And this looks bad for the vaccination because it’s much higher…

But - No. It doesn’t tell us anything, it’s too big an age group. In that age group 10-59 the people most likely to die are the 50-59s, and they’re also the ones more likely to be vaccinated. People have picked up on this as evidence the vaccine is terrible…but it doesn’t show anything.

{ED comment: Point well made, but as the highest age groups are excluded one wonders if the effect of this counfounding-by-age factor will be sufficient to completely negate the point made by the graph, as 50-59 isn’t exactly Methuselah… One also wonders why the ONS would bundle the 10-59 together in the first place! Very odd. Was it to conceal anything to do with younger age groups? Bear in mind that these graphs only appeared after an FoI from the Fenton team!}

DB points out that “If you had an agenda you would just throw this out there - but you want to get this right, so you explain the anomaly, the unvaccinated are younger, they’re healthier, so you take this a step further, let’s get to that.”

NF Let’s start looking at the older age groups.

DB points out that that peak is spectacular. “Did a volcano go off!?”

NF: But there’s something wrong with this graph. It peaks at the same time as the first vaccine roll out for this age group peaks.

And you see a similar pattern, with the peak in a different place, for the other age groups. Look at the 70-79:

They were vaccinated earlier, that’s where you see the peak of mortality.

Strangely, again it peaks at the same time as the first vaccine roll out for this age group peaks.

Again, go to the 80+ group:

Same again: the peak moves back to the peak of the vaccine rollout for the age group.

DB: So they’re all having this peak in the unvaccinated, at the exact…at different times of the year, so you can’t blame something else - at the time the vaccine started. They’re all having this peak in the unvaccinated - who didn’t receive the vaccine!

NF: Exactly. It’s crazy. Now people have said {gestures as if unimpressed} maybe covid mortality caused these figures.

On the next graph, you’ll see this is not the case.

NF: These are the non-covid deaths - they have nothing to do with covid. You’ve got this bizarre situation, somehow when the vaccine rollout happens, at its peak, the unvaccinated are dying of something unrelated to covid!? And they’re dying at a much greater rate than the vaccinated, it’s ridiculous.

DB: {Addresses the audience}: If you’re vaccinating a large group of people, you would expect, if there are injuries from the vaccine, for them to show a little peak - but why would the other group, you know, standing on the other side of the fence, you know, across the field, nothing to do with this vaccine - why would THEY be having a peak in death at the same moment the people getting the vaccines? Unless, {they both speak} - professor?

NF: …misclassifications…here is a hypothetical example of how a very simple misclassification

(DB comes in: you know what this … - they both seem to be smiling knowingly…let’s get to the example)

NF agrees

In this example I’m going to do something other than a misclassification. A vaccine in a population of 10m people. But actually, it’s not really a vaccine, but a placebo, with absolutely no effect.

You can see, it doesn’t decrease or increase deaths. And if you assume that the mortality rate in this popn is 50 per 100,000, then every week, you would expect to see per hundred thousand, fifty people die.

{ED comment: Up to week 8 the vaccination rate is being ramped up, doubling each week}

Here’s the little trick. All I’m going to do now, is just suppose there is a delay of one week in the death reporting. Let’s suppose that 50 deaths in week 1 gets reported as the number of deaths in week 2, and so on.

And look what happens to the mortality rates. I’ve done nothing other than shift the death reporting by one week. Somehow, the vaccinated, are…less than half the mortality of the unvaccinated, even though they’re all getting the same placebo.

And look at the unvaxxed stats - the mortality starts to really peak, at the time that the vaccine rollout has peaked.

DB Wow - and if you graph that out, what does it look like

NF:
image

DB: There we are, you have a perfect peak that looks just like the peak that we’re actually seeing in reality.

NF: Exactly. This placebo - we know it’s a placebo, because we set it up to be - is now a miracle vaccine, it’s saving people.

I should say that we think the death reporting in the ONS report wasn’t delayed by a week, BUT - if someone dies shortly after a vaccination, a week or a couple of weeks later, and if it’s not reported as being vaccinated - as we believe is likely, even though they say they will classify someone as having a vaccination within 21 days, even if it’s a day later they will still count that person as having one jab…

DB: Let’s be clear, and I don’t know if it’s the same in England, but in America, our CDC tells us we’re not going to consider you vaccinated until 14 days. This is on our website:

{CDC page is When You’ve Been Fully Vaccinated | CDC}

You have to be 2 weeks after your second dose in a two-dose series (Pfizer and Moderna) so that month after your first shot you are considered unvaccinated if you get sick and die you’re going to land, categorised as unvaccinated. And for the JJ vaccine you have to wait two weeks after that single dose.

That there could create this exact same anomaly too

NF: It could, it would - if you just did it by one week you would get (speaks assertively) identical figures to what I’ve shown, and you can prove it’s identical, my paper - (indistinct, mentions blog) …

To be fair, I have to say, the ONS is telling us, that we don’t do that 14 days, that even if it’s less than 14 days, if they’ve died they’re supposed to be recorded as vaccinated, but the way things work in hospitals and surgeries, if they die shortly afterwards - the paperwork isn’t going to be submitted. It seems to me unlikely that the information on when they got vaccinated is going to be there.

DB Especially in a situation where a person goes to a clinic, or their personal doctor, to get vaccinated. If they end up in the ER, or dying, they’re in the nearest hospital, they could be in a different system. Many times what we’re seeing is that you’re only listed as having been vaccinated if you’re vaccinated in the same hospital you end up in for your critical care, so right there you can see problems and anomalies.

NF Exactly…

DB Next slide?

NF Yes

These plots are three different classes of vaccination. These are people who are vaccinated.

These are non-covid deaths, and are different categories of vaccinated people.

These should all be the same, and we’ve looked at the life mortality table

{ED: Del makes the thick grey line (Lifetable mortality data) come up on the graph - as this is non-covid deaths, ALL the lines should follow this!!}

{This is similar to Figure 11 on the paper, which explains that

" The vaccinated are categorised into three different categories, namely: ‘within 21 days of first dose’, ‘at least 21 days after first dose’, and ‘second dose’. However, in each age category the mortality fluctuates in a wild, but consistent way."

Figure 11 also confirms that the ‘at least 21 days’ IS the one with the high mortality - NF says on the clip he is colour blind, so I went and checked

They chat to the effect that we should be seeing the grey line for all three!}

NF continues: You get the second dose and suddenly you’re immortal - this is a cure for all medical conditions!

DB: Remember this is not covid deaths, the vaccine shouldn’t be affecting any of it.

NF We don’t know why this data is so screwed up, we have some hyptotheses.

Here is another anomaly {This is a brief digression}

I’m not saying this changes anything, but why are they doing this?

{Prof NF rails a bit more…}

It just shows you how unreliable the data is. So what we’ve got to do is hypothesize what the reasons are

for these anomalies in the data, and how we can adjust them.

{ED: that was a digression, back to the main now}

NF: What we’ve done here, is we DO the adjustment where we make the assumption that the miscategorisation…

(DB interrupts to say what the dotted lines are - NF explains these just show when the vaccine rollouts are)

When the first dose was being rolled out, when we’ve done the adjustments for the mis-categorisation and also adjusted for the lifetime mortality figures, ie adjusted for the difference in the life mortality figures

{ED: This graph is Figure 16 of the linked paper. See the concise explanation above Figure 16 on P12. They adjust the unvaccinated data for the implausible peak after the dose 1 rollout peak, and re-allocate to the first dose and second dose mortality each week.

The paper explains " Hence, excess mortality in the unvaccinated was assigned to the single dose vaccinated and that in the single dose vaccinated was assigned to the double dosed."

Hence the elevation in the vaccinated deaths; the spike at dose 1

Paper: " In all cases the spike begins with the roll out of the first dose for each age group." }

NF It’s all non-covid mortality. So you’ve got this peak. Now you’re seeing the vaccinated people dying at a higher rate at the beginning of the period where the vaccination is rolled out for that group.

Possible explanation: in the UK they prioritized it for each age group. Those with most co-morbidities first, the particularly frail. Because those people were genuinely the most at risk of dying from covid, it made sense to vaccinate those people first. But of course what we’re seeing here, is that it appears those most frail people - because they get that quick hit immediately the vaccination is rolled out, a possible explanation, we’re not saying we’re sure, it’s a reasonable explanation, those people not only at most risk of covid, but of death of the vaccine. The probability is, those are the people who probably would have died within a few week anyway, the vaccine just kind of pushed them over the edge. There is no other - we looked at the other possible explanations, socio-economic factors, the paper looks at a whole lot of stuff.

We’ve had people trying to discredit us on this, because they offer a different explanation. And do you know what their explanation is? They say actually the original graphs are correct, it is the unvaccinated who are dying at the time they are rolling out the vaccine, they are saying that because these people are close to death, they are chosen not to have the vaccine.

Now we know that is not true in the UK. We know people close to death were getting the vaccine.

DB: I see, it was the same in the US - if you’ve got heart disease, cancer - you gotta get the vaccine.

Let’s go to your conclusions

NF: OK

Conclusions

Vaccine ‘effectiveness’ studies are generally flawed because of the reliance on determining covid-19 ‘cases’.

The simplest and most objective way of determining the overall risk-benefit of the covid-19 vaccines is to compare all-cause mortality between the vaccinated and the unvaccinated.

Latest ONS mortality report by vaccine status should provide the necessary data but is flawed in many ways.

The anomalies most easily explained by misclassification of some vaccinated deaths as unvaccinated.

After adjusting for the misclassification it appears that shortly after vaccination people may be exposed to an increased mortality risk.

Whatever the explanations the ONS data is both unreliable and misleading. Absent any better reason, Occam’s razor would support our conclusions.

The ONS data provide no reliable evidence that the vaccines reduce all-cause mortality .

Del: So at the very least, you can not be using this data to say, Look how great our vaccines are, you unvaccinated are at higher risk - we know it, because of our data.

This data, when it’s properly looked at, shows that there is a problem that is not making sense and appears to be caused by the vaccine itself.

NF: And what it definitely does not do is provide any data that proves the vaccine is particularly safe and effective. It doesn’t provide the evidence for that.

And in a sense we should not be surprised, one of the slides we skipped through, for the Pfizer - the randomized controlled trial,

DB: here:
image

{ED: what’s featured here is that there was more all-cause mortality in the vaxed group than the placebo group:}
image

More chat, and DB’s very warm thanks and praise for Prof Norman Fenton and his team.

Link again: MATHEMATICIAN UNCOVERS DISTURBING MORTALITY DATA - The HighWire

3 Likes

Hi @Evvy_dense , great work thanks. I’ll need to read it over a few times to fix all the pieces in place.

I see the paper was authored mostly by QM college London, I bet BG hasn’t been stuffing funds into their research like he has with UC London the home of the notorious Ferguson!

Cheers

1 Like

Seconded.

I watched the program and couldn’t believe … well actually I could believe it, and it’s exactly what some of us have come to expect from the deep state and that “reliable” source of data the ONS.

Anyway, I’m off to get me a placebo to save me from whatever it is.

2 Likes

Yes, thanks for the exegesis, Evvy. Painstakingly done.

2 Likes

Many thanks @Evvy_dense for all your effort.

The age-related roll-out - as Prof Fenton says - made sense on health grounds. It also provided some very handy reference points for analysing the data early on. Was there a logic to creating that so very wide 10-59 age bracket (I am 59 btw)?

Was there a logic to 10-59? I think so :wink:. It only appeared when they were FoI’d by Fenton’s team, so they had to ‘publish’ the data. The outcome-reversing anomaly that age creates when it’s not broken down (and 10-59 is not a breakdown) is well known, they’ve been fielding it since the start when all the vaxed were pensioners so obviously the vaxed died in greater numbers anyway until the vaccination is spread evenly in the population. I think they knew this could be fielded by the usual reliable people. Of course the anomaly would have disappeared if they had presented ten-year age groups. It’s quite possible something worse would have appeared in the younger age groups whose data is being thereby hidden.
You’ll be released from the lied-about data set next year, to new statistical adventures!

1 Like

That was my recollection of following these ONS figures throughout the year: just a hella nice big bucket.