Bayesian analysis is the right framework for settling the COVID-19 origins question

This is still a viable conspiracy theory that could be true. As far as I'm concerned, it's more likely to be true than the wet market origins. So, I think that's why, again during COVID, what I would have liked to see is our public policymakers and p...

You can watch this play out on genuinely unresolved questions like COVID's origins, where the honest position is an explicit probability that shifts as withheld evidence becomes available — not a flag planted on one side.

The core Bayesian formula. It allows us to calculate what our hypothesis H should be given the evidence E. The only problem with this formula is that it can not easily integrate multiple pieces of evidence. For that we will need to do a slight rewrit...

Here, we introduce KM-GPT-DCH, an algorithm that combines co-occurrence methods with large language models (LLMs) to develop a transparent and reproducible literature-based algorithm to compare controversial hypotheses using a structured scoring appr...

Throughout the exchange I returned repeatedly to the same request: assign probabilities to the six propositions and multiply them together.

Seems to me a complete Bayesian treatment would require explicit probability assessments, or at least sensitivity tests, for the full sequence [...] I can't see how a full Bayesian analysis wouldn't integrate, or at least sensitivity-test, these beha...

After correcting a fundamental error in Bayesian reasoning, the results presented in that paper imply larger likelihood for a single introduction, reducing the plausibility of the wet-market zoonosis account of Covid’s origins.

Using only the data, model, and simulations of P2022 itself, I show that once a basic logic error is fixed the P2022 approach implies that a single successful spillover of SARS CoV-2 was more likely than a double successful spillover pair. The EJW pa...

Weissman uses a simple analogy to explain the error. Correcting the error using the data, model, and simulations of the original paper reverses the implication of the analysis-the single-introduction likelihood becomes greater than the two-introducti...

Ridley described his argument as a "good old Bayesian thing, you add up all the probabilities until you get to a final result," or a court case in which a jury reaches the most reasonable conclusion.

Instead, Ridley described his argument as a "good old Bayesian thing, you add up all the probabilities until you get to a final result," or a court case in which a jury reaches the most reasonable conclusion.

There are two types of evidence, broadly speaking, we can bring to bear on this question. The first is epidemiological – looking at the earliest cases of COVID including the strains of virus with which they were infected, and data on the virus (such ...

An influential article claimed that Bayesian analysis of the molecular phylogeny of early SARS-CoV-2 cases indicated that the likelihood that two successful introductions to humans had occurred was greater than the likelihood that just one had occurr...

The Math Doesn’t Lie: A Bayesian Analysis of COVID-19’s Origin

Shermer uses Bayesian reasoning to distinguish real conspiracies from speculation—showing why even die-hard skeptics must update their views when hard evidence emerges.

Examples from the pandemic—the debate over covid origins and expert guidance on public health measures—will be used to (i) illustrate how Bayesian inference works, (ii) highlight shortcomings in expert reasoning, and (iii) call attention to the poten...

Bayesian probability provides a natural and intuitive framework for characterising and communicating uncertainty. Bayesian analysis simply applies the laws of probability to evaluate which hypothesis is more plausible in light of whatever relevant in...

Bayesian Phylogeography methods were used via the BEAST software. The first case of COVID-19 in Uganda likely originated from Kenya.

The study utilised genomic sequences from the Global Initiative on Sharing Avian Influenza Data (GISAID) repository. Bayesian Phylogeography methods were used via the BEAST software. The first case of COVID-19 in Uganda likely originated from Kenya.

Bayesian probability provides a natural and intuitive framework for characterising and communicating uncertainty. Bayesian analysis simply applies the laws of probability to evaluate which hypothesis is more plausible in light of whatever relevant in...

How would they respond to my suggestion that “false confidence” is not a problem and that belief-as-probability is enough to deal with satellite collision avoidance?

A coronavirus outbreak in Wuhan is more surprising given a natural zoonosis event than it is given a lab escape event. From a Bayesian standpoint, the outbreak location favors the lab escape scenario.

Using a log-normal model to allow for a fourfold uncertainty in early counts, the data are about 10 times more likely under the market-origin scenario than the lab-origin scenario. My calculation was a deliberately simple likelihood comparison: given...

Bayesian probability is about quantifying a "state of information". Retrospectively, that is, not what did happen, but rather how much you know about what did happen. [...] from what we know, we need to decide which of the possible explanations has t...

What are the chances this is unconnected with the pandemic outbreak? Weissman attempted to quantify this but it is clear that the probability must be vanishingly small.

The conclusion that one should reach is to assign equal probabilities to all hypothesis and allow the evidence to decide which is more likely.

Infectious clones made via reverse genetics systems have featured prominently in my own work on SARS-CoV-2 origins for a reason. Behind my colleagues’ and my 2022 paper quantifying the magnitude of an anomaly in the SARS-CoV-2 genome consistent with ...

A higher probability of a lab-leak should also reduce the probability of zoonotic origin but the latter is an already known risk and COVID doesn’t add much to our prior while the former is new and so the net probability is positive.

Instead of a Bayes factor of 0.25, we get a Bayes factor of 11.95 because one reasonable model (early cases predict locations of subsequent cases) is compared to an unreasonable model (everyone is infected by raccoon dogs).

Bayesian methods have grown in popularity in the past few decades with the advent of improved statistical software and computing power. But Levin’s paper is the first to apply the technique to the COVID origins debate.

Bayesian Reasoning is a way to believe true things. It is a method to update one’s beliefs given some evidence, so that one ends up with more credence on beliefs that match the evidence. While Bayesian Reasoning is not the only method to find true co...

The lab-leak theory should thus have topped every initial differential diagnosis and every list of subjective "prior probabilities."

A fundamental objective of this study has been the formulation of a Bayesian framework that can be used to assess the relative weight of the evidence for the two key hypotheses, that is, accidental lab leak vs. zoonotic spillover.

Thanks for the great analysis! One element that I don’t find entirely convincing is the part about considering the uncertainty in the log of the prior odds, e.g. assuming a 3-df-t-distribution with an SD of 2.3. What is missing is an upper bound on t...

Bayesian inference, a method of statistical inference based on Bayes' Theorem, enables calculation of probabilities of hypotheses based on multiple lines of evidence and enables updating of probabilities of hypotheses as additional lines of evidence ...

Don’t give up so easily! We as a society have spent approximately 0% of our collective decision-making resources on explicit Bayesian reasoning. Just because Rootclaim used Bayesian methods and then lost a debate doesn’t mean those methods will never...

Whatever restriction you choose to apply to the virus on the lab-leak side, you need to apply to zoonosis (otherwise you’re calculating conditional probabilities of different evidence for each side).

Multiplying the share of markets (1 and 2) by the share of HSM within markets (3), we converge to around 3-5%.

Weissman's Bayesian analysis provides a thorough overview and is probably as good a case for lab origin as any.

The Superforecasters frequently referenced base rates, i.e., the history of pandemics emerging from natural zoonosis versus the history of laboratory leaks, to anchor their probabilities.

My point was that in some cases the update can be so strong that it overrides all reasonable uncertainties in priors and personal estimates. And exactly this makes Bayes’ theorem useful and strong instrument.

We therefore never provided extremely low conditional probabilities under zoonosis, and as a result didn’t have any extreme factors in our analysis. Unfortunately, the result of our steelmanning was that when our hypothesis’ explanation was favored, ...

It's hard for me to objectively ballpark (4) without help from a few experts. But (1), (2), and (3) are fairly easy to get a ballpark figure for. I think these three are all pretty much independent, so the overall probability of these three things ha...

What we do have, however, is Bayes’ powerful theorem, an established method for calculating the odds of any particular outcome by factoring in the events that are credibly related to that outcome. It is widely used in science—although apparently many...

We’ve seen that calculating Bayesian odds involves both picking priors and calculating likelihood ratios.

Bayesian reasoning serves as a bridge between quantitative and qualitative research that allows one to analyze any and all kinds of data within a single, unified inferential framework. This paper demonstrates the strength and flexibility of the Bayes...

It’s worth noting that rootclaim.com got this right. They have many bold claims and they seem to be using good Bayesian reasoning.

My priors: Lab leak 60% Natural origin 40%

I thought the whole point of Bayesianism was to make more clear the often arbitrary and/or questionable priors that we all start research with.

A Bayesian analysis concludes beyond a reasonable doubt that SARS-CoV-2 is not a natural zoonosis but instead is laboratory derived.

This post has not attempted to derive the one true and absolute probability that the first outbreak of SARS-CoV-2 in Wuhan originated from an uncontrolled release from a laboratory. It has presented a Bayesian analysis, with sceptical priors, and a p...

Given that full transparency from Chinese authorities is unlikely, assessing the probabilities is the best we can do. Fortunately, that has been done with with impressive scientific rigour by DRASTIC member Dr. Steven Quay MD, PhD in his technically ...

Waterfalls are one way of visualizing the odds form of Bayes’ rule, which states that the prior odds times the likelihood ratio equals the posterior odds.

Here, we introduce KM-GPT-DCH, an algorithm that combines co-occurrence methods with large language models (LLMs) to develop a transparent and reproducible literature-based algorithm to compare controversial hypotheses using a structured scoring appr...

Here, we introduce KM-GPT-DCH, an algorithm that combines co-occurrence methods with large language models (LLMs) to develop a transparent and reproducible literature-based algorithm to compare controversial hypotheses using a structured scoring appr...

Here, we introduce KM-GPT-DCH, an algorithm that combines co-occurrence methods with large language models (LLMs) to develop a transparent and reproducible literature-based algorithm to compare controversial hypotheses using a structured scoring appr...

Here, we introduce KM-GPT-DCH, an algorithm that combines co-occurrence methods with large language models (LLMs) to develop a transparent and reproducible literature-based algorithm to compare controversial hypotheses using a structured scoring appr...

Here, we introduce KM-GPT-DCH, an algorithm that combines co-occurrence methods with large language models (LLMs) to develop a transparent and reproducible literature-based algorithm to compare controversial hypotheses using a structured scoring appr...

After correcting a fundamental error in Bayesian reasoning, the results presented in that paper imply larger likelihood for a single introduction, reducing the plausibility of the wet-market zoonosis account of Covid’s origins.

The study utilised genomic sequences from the Global Initiative on Sharing Avian Influenza Data (GISAID) repository. Bayesian Phylogeography methods were used via the BEAST software. The first case of COVID-19 in Uganda likely originated from Kenya.

The study utilised genomic sequences from the Global Initiative on Sharing Avian Influenza Data (GISAID) repository. Bayesian Phylogeography methods were used via the BEAST software. The first case of COVID-19 in Uganda likely originated from Kenya.

The study utilised genomic sequences from the Global Initiative on Sharing Avian Influenza Data (GISAID) repository. Bayesian Phylogeography methods were used via the BEAST software. The first case of COVID-19 in Uganda likely originated from Kenya.

The study utilised genomic sequences from the Global Initiative on Sharing Avian Influenza Data (GISAID) repository. Bayesian Phylogeography methods were used via the BEAST software. The first case of COVID-19 in Uganda likely originated from Kenya.

Bayesian probability provides a natural and intuitive framework for characterising and communicating uncertainty. Bayesian analysis simply applies the laws of probability to evaluate which hypothesis is more plausible in light of whatever relevant in...

Bayesian probability provides a natural and intuitive framework for characterising and communicating uncertainty. Bayesian analysis simply applies the laws of probability to evaluate which hypothesis is more plausible in light of whatever relevant in...

Bayesian probability provides a natural and intuitive framework for characterising and communicating uncertainty. Bayesian analysis simply applies the laws of probability to evaluate which hypothesis is more plausible in light of whatever relevant in...

Waterfalls are one way of visualizing the odds form of Bayes’ rule, which states that the prior odds times the likelihood ratio equals the posterior odds.

The arguments around covid origins seem largely to fall into an inference trap Andrew has termed premature collapse of the wavefunction. Rather than maintaining multiple hypotheses with different levels of credibility, there seems to be a desire to c...

In the 78-page document, we determined that most of the peer-reviewed scientific evidence supports the hypothesis that SARS-CoV-2 has a zoonotic origin, meaning that it came from an animal.

In the 78-page document, we determined that most of the peer-reviewed scientific evidence supports the hypothesis that SARS-CoV-2 has a zoonotic origin, meaning that it came from an animal. But until requests for additional information are met or mor...

In the 78-page document, we determined that most of the peer-reviewed scientific evidence supports the hypothesis that SARS-CoV-2 has a zoonotic origin, meaning that it came from an animal. But until requests for additional information are met or mor...

In the 78-page document, we determined that most of the peer-reviewed scientific evidence supports the hypothesis that SARS-CoV-2 has a zoonotic origin, meaning that it came from an animal. But until requests for additional information are met or mor...

In the 78-page document, we determined that most of the peer-reviewed scientific evidence supports the hypothesis that SARS-CoV-2 has a zoonotic origin, meaning that it came from an animal. But until requests for additional information are met or mor...

The origins of the coronavirus pandemic remain contested; the evidence is incomplete.

The origin of COVID is still uncertain.

The origins of COVID-19 remain uncertain. A lab-associated incident is plausible; so is natural spillover. China’s obstruction has made the truth harder to establish, and scientific opinions vary.

The origins of COVID-19 remain uncertain. A lab-associated incident is plausible; so is natural spillover.

In my post “Why I’m not a Bayesian”, I argued that the Bayesian approach of assigning credences to propositions with binary truth values only works in simple and restricted domains. Instead, I claimed, a better approach to epistemology is to assign d...

Six independent Bayesian analyses of the same evidence spanned 23 orders of magnitude. [...] all this information is still incredibly difficult to navigate, interrogate, and use to inform one’s beliefs.

Over 15 hours of structured argument, two smart people marshalled epidemiological data, viral genetics, Bayesian inference, and institutional analysis to reach opposite conclusions. Two expert judges ruled decisively for zoonosis. Six independent Bay...

The COVID origin analysis was an exercise in anchor bias. Instead of gathering data, drawing conclusions, and writing multiple products based on the data, the analysts worked from a binary choice: lab incident or natural origin.

This post is like insisting that earth is flat because you and some friends took some photos on which the horizon looks flat, then insisting that others should run a bayesian analysis based only on these 5 pieces of easily manipulatable evidence.

Weissman prefers a Bayesian approach, where you average over counterfactual realities, where one detail could have been different. [...] This process of averaging over different possibilities can go on forever, and no one agrees on which counterfactu...

The competing hypotheses on the origins of SARS-CoV-2 – the pathogen responsible for the COVID-19 pandemic – have been hotly debated in the scientific literature and popular media, especially with respect to the hypothesis that SARS-Cov-2 originated ...

In the 78-page document, we determined that most of the peer-reviewed scientific evidence supports the hypothesis that SARS-CoV-2 has a zoonotic origin, meaning that it came from an animal.

While available information is insufficient to definitively conclude the origins of SARS-CoV-2, the independent assessment of SARS-CoV-2 origins report (June 2025) published by the Scientific Advisory Group for the Origins of Novel Pathogens (SAGO) r...

The former case is cause to relax; the latter is not. Yeah, in a sense this is obvious. But it's a reminder that seeing a probability estimate isn't a substitute for real diligence.

In Rootclaim’s most recent COVID origins analysis, the single strongest piece of evidence is “12 nucleotides clean insertion,” which they claim is 20x more likely in lab leak (after out-of-model correction). [...] So, they do not have any evidence th...

The take away is just that if you want to predict rare events, you need evidence. Rare events do happen, if rarely, and unless you collect the evidence you'll have no choice but to be surprised if it happens to you.

Science alone cannot tell us how Covid emerged in Wuhan in 2019. An investigation must determine whether the virus was studied in a laboratory before the detected outbreak. Intelligence collected from human sources, intercepted signals or data, and o...

Without China providing fundamental data, definitive conclusions remain impossible.

Without China providing fundamental data, definitive conclusions remain impossible.

Early viral genome sequences remain unreleased, and environmental data from Wuhan’s Huanan Seafood Market were shared only after international researchers briefly gained access online.

We have no idea who patient zero was or where that person contracted the virus. Everyone goes to food markets or has someone in their household who goes to food markets, so all this downstream analysis attempting to locate patient zero in physical sp...

One of the senior authors presents the likelihood of a single introduction failing to reproduce observations as the probability that there were two introductions. Pekar does not hint at non-independent introductions, and explicitly estimates the numb...

In my view the answer here is so obvious that a formal analysis is not really necessary. There is basically no chance at all that this is a coincidence.

I have a very incomplete understanding of statistics. I understand sort of how to use probability prospectively. Retrospectively has me confused. The Washington Nationals are unlikely to win a World Series; therefore 2019 did not happen?

It remains a priority to study the origins of COVID. While the evidence for accidental origin of COVID is not definitive, there is a substantial body of strong circumstantial evidence. US intelligence agencies remain divided. The WHO Director has sai...

A recent World Health Organization report presents the newest findings on the origins of COVID-19 from the Scientific Advisory Group for the Origins of Novel Pathogens (SAGO), an international body of more than 24 independent scientists. Drawing from...

Both sides are gathering evidence to support their case, yet neither can fully rule out the possibility put forward by the other.

I also demonstrate clear evidence of bias among the earliest available sequences, and explain why the root of the SARS-CoV-2 tree cannot be conclusively resolved with the current data.

You get the feeling that they often start from the conclusion and develop their arguments backward.

They can do their probabilities, they can dig up their documents, they can try to bang square pegs into a round hole, but what they can't get right is the virus stuff.

But how can you prove anything when you don't have access to the WIV laboratories? It's just highly suspicious.

Several scientists thought that the origin of the coronavirus had not been sufficiently investigated.

I guess we are counting on you as a Bayesian expert to eventually come up with an opinion. I think your “composite hypothesis” criticism is very telling: a good way to stack the deck.

But I don’t see how you can do that, and lay claim to a probabilistic assessment, without even attempting to quantify the probabilities that only the experts you agree with are trustworthy (stars even!) while the majority of scientists with relevant ...

In response, co-author Washburne resorts to special pleading about maximum fragment size being a relevant statistic only for engineered viruses. But at the same time, this special pleading was not used to address that there is that tiny fragment that...

It’s easy to fool oneself into thinking one has a solid argument based on math when one really just has a “mathy” argument that depends on shaky premises.

I agree with this! It seems like a great cover for motivated reasoning, use some Bayesian analysis to come to your preferred conclusion. What is missing here is any input from relevant experts (virologists, genomics, epidemiologists, etc) to inform t...

In fact I think this is a problem for public understanding of science [...] dressing up pseudoscience with a veneer of “sciency”-sounding arguments [...].

It's hard to compute Bayesian probabilities for this problem, not so much because of the priors but because of the likelihood, which is the probability of the data given the model. One problem is the selection of what is considered to be data, the ot...

Anything claiming to be “Bayesian” but not doing actual Bayesian statistics (fitting models to data and then doing an inference from a posterior distribution, justifying all steps especially the prior) is untrustworthy. I have never seen a convincing...

The agency continues to assess that both research-related and natural origin scenarios of the COVID-19 pandemic remain plausible.

You cannot apply Bayes' Theorem until you have a probability space; many real-world situations, especially the ones people argue about, do not have well-defined probability spaces, including a complete set of mutually exclusive and exhaustive possibl...

Both look at the same data set. Both use Bayesian reasoning. One says 100,000,000-1 lab leak and one says somewhere between 10,000-1 and 10,000,000-1 zoonosis. [...] even with this sort of methodological agreement, they end up about as far apart as t...

In view of the difficulty in establishing a posterior odds ratio for the number of spillovers, it appears that the evidence for this narrative is still inconclusive.

In view of the difficulty in establishing a posterior odds ratio for the number of spillovers, it appears that the evidence for this narrative is still inconclusive.

On the Rootclaim thing: My impression was that they were doing Bayes badly. I have to admit I got a bit bogged down in the incredibly long article about it and all the back-and-forth, but as I understood it they really felt that you could come up wit...

Determining the origin of an outbreak or pandemic can be divided into four stages: detection, identification, characterization, and attribution. Understanding how a specific pathogen entered and spread in a population to cause an outbreak is a multid...

I am skeptical that the Bayesian decision making/evaluation methods are any more "objective" than [intuitive reasoning]. I think they maximize legibility, not objectivity, and tend to hide the intuitive/heuristic portion in the data inclusion step an...

six independent Bayesian analyses of the same evidence spanned 23 orders of magnitude [...] demonstrating that Rootclaim's 'heroic Bayesian analysis' cannot resolve disagreements - too much evidence, too hard to quantify.

Still, this is faint praise for a method that hoped to be able to resolve these kinds of disagreements.

I was concerned that people might interpret my Bayesian analysis as the “main product” of the report, in isolation from the rest, which is why I heavily cautioned against taking it too literally; indeed I explicitly state that I do not think it is an...

Now reducing human biases and providing aid to reasoning are certainly lofty goals one can and ought to agree with. But the devil is, as often, in the details and implementation. How it works: Rootclaim defines basically a set of hypotheses, then tak...

To be clear, my problem is not with the possibility of a lab leak itself, it’s with the evaluation that present evidences are anything but posthoc rationalizations fueled by unhealthy levels of tunnel vision.

That's a good reductio for taking even fairly well justified sounding subjective bayes factors at face value.

That is, this type of informal Bayesian estimate is useful as part of a ritual for changing your own mind, when done carefully. That requires a significant degree of self-composure, a willingness to change one's mind, and a high degree of justified c...

scientifically, neither our statistical analysis nor that of Worobey et al. could reject or support either the natural origin or lab-leak theory.

Both sides attempted Bayesian estimates and probabilities and got absolutely absurd differences in estimates. Rootclaim failed to impress me - the takeaway I got is that they are well suited to say murder cases where there is history to go off, but w...

Bayesian analysis…you can’t be serious. This isn’t science, it’s business.

The value add by this “bayesian analysis” is hence 0. It’s just numbers and keywords thrown in for a sense of scientific authority for what is an entirely qualitative argument.

If the bayesian stuff makes no sense, that doesn’t bode well for the rest of the paper.

In the 78-page document, we determined that most of the peer-reviewed scientific evidence supports the hypothesis that SARS-CoV-2 has a zoonotic origin, meaning that it came from an animal.

In the 78-page document, we determined that most of the peer-reviewed scientific evidence supports the hypothesis that SARS-CoV-2 has a zoonotic origin, meaning that it came from an animal.

In the 78-page document, we determined that most of the peer-reviewed scientific evidence supports the hypothesis that SARS-CoV-2 has a zoonotic origin, meaning that it came from an animal.

In the 78-page document, we determined that most of the peer-reviewed scientific evidence supports the hypothesis that SARS-CoV-2 has a zoonotic origin, meaning that it came from an animal.

In the 78-page document, we determined that most of the peer-reviewed scientific evidence supports the hypothesis that SARS-CoV-2 has a zoonotic origin, meaning that it came from an animal. But until requests for additional information are met or mor...

In the 78-page document, we determined that most of the peer-reviewed scientific evidence supports the hypothesis that SARS-CoV-2 has a zoonotic origin, meaning that it came from an animal. But until requests for additional information are met or mor...

In the 78-page document, we determined that most of the peer-reviewed scientific evidence supports the hypothesis that SARS-CoV-2 has a zoonotic origin, meaning that it came from an animal. But until requests for additional information are met or mor...