@bratwebb.bsky.social
I am suggesting that Chinese data sharing shows 92 potential early cases that were eliminated. It seems that you are suggesting that China was honest about the 92 initially and dishonest about the later serology tests that eliminated them later.
December 31, 2025 at 6:03 PM
I am suggesting that Chinese data sharing shows 92 potential early cases that were eliminated. It seems that you are suggesting that China was honest about the 92 initially and dishonest about the later serology tests that eliminated them later.
Identification from Chinese data, rejection from Chinese data
December 31, 2025 at 5:52 PM
Identification from Chinese data, rejection from Chinese data
That 92 were considered plausible is also from Chinese data.
December 31, 2025 at 5:50 PM
That 92 were considered plausible is also from Chinese data.
The 92 that were eliminated by WHO
December 31, 2025 at 5:38 PM
The 92 that were eliminated by WHO
Reference confirming the 266 count and any implied November cases that were eliminated by WHO: archive.vn/2WSIi#select...
archive.vn
December 31, 2025 at 5:36 PM
Reference confirming the 266 count and any implied November cases that were eliminated by WHO: archive.vn/2WSIi#select...
AFAIK, SCMP said 9 cases in November and 266 overall. This is consistent with the 174 early cases from December and 92 that the WHO considered and eliminated, so the SCMP November case reference is invalid.
December 31, 2025 at 5:30 PM
AFAIK, SCMP said 9 cases in November and 266 overall. This is consistent with the 174 early cases from December and 92 that the WHO considered and eliminated, so the SCMP November case reference is invalid.
The short window is relevant in the context of multiple spillovers reducing the D=2 penalty. As I've noted, 'statistical inference without biological context' reminds me of several climate science debates disconnected from domain-specific knowledge that neither of us possesses.
December 17, 2025 at 6:55 PM
The short window is relevant in the context of multiple spillovers reducing the D=2 penalty. As I've noted, 'statistical inference without biological context' reminds me of several climate science debates disconnected from domain-specific knowledge that neither of us possesses.
In summary, reservoir modeling is complex. D=0/1 penalties are greatly reduced with multiple spillovers from a reservoir that does not include TT/CC for biological/sampling reasons. A short window correlated introduction raises the probability that A/B are established in close time/space proximity.
December 17, 2025 at 6:17 PM
In summary, reservoir modeling is complex. D=0/1 penalties are greatly reduced with multiple spillovers from a reservoir that does not include TT/CC for biological/sampling reasons. A short window correlated introduction raises the probability that A/B are established in close time/space proximity.
Separate cages etc may also eliminate intermediates in the reservoir or they may be gapped release not because of fitness but sampling bias upon bringing animals to the market. Hensel '25 example supports that there may be bottlenecks for intermediates.
December 17, 2025 at 5:56 PM
Separate cages etc may also eliminate intermediates in the reservoir or they may be gapped release not because of fitness but sampling bias upon bringing animals to the market. Hensel '25 example supports that there may be bottlenecks for intermediates.
Your reservoir penalty of 0.27 (Poisson) for D=2 includes possibilities that D=0/1. In a short time window, with correlated intros, multiple spillovers of A/A+1 and B would be indistinguishable from A/B satisfying time constraints. But TT/CC is not cryptic and may be selected against like in humans.
December 17, 2025 at 5:48 PM
Your reservoir penalty of 0.27 (Poisson) for D=2 includes possibilities that D=0/1. In a short time window, with correlated intros, multiple spillovers of A/A+1 and B would be indistinguishable from A/B satisfying time constraints. But TT/CC is not cryptic and may be selected against like in humans.
I'm saying host selection etc against certain mutations may create gapped reservoirs. A+1/B+1 are not D=1 penalties for multiple (cryptic) spillovers in a short time window correlated introduction environment but TT/CC are and they may be selected against.
December 17, 2025 at 4:49 PM
I'm saying host selection etc against certain mutations may create gapped reservoirs. A+1/B+1 are not D=1 penalties for multiple (cryptic) spillovers in a short time window correlated introduction environment but TT/CC are and they may be selected against.
Mutation patterns in animals could be different and favor pre-existing diversity. From Hensel '25 "Multiple spillovers need not have high sequence diversity—for instance, the viral genomes RshSTT182 and RshSTT200 were collected from two different bats and differ by only 3 mutations"
December 17, 2025 at 3:44 PM
Mutation patterns in animals could be different and favor pre-existing diversity. From Hensel '25 "Multiple spillovers need not have high sequence diversity—for instance, the viral genomes RshSTT182 and RshSTT200 were collected from two different bats and differ by only 3 mutations"
As long as the animal reservoir is a well-mixed population that can be modeled by the Poisson distribution from which D=2 is drawn, this seems to be correct by my understanding. I don't know if that is biologically a good model though as opposed to an I1 evolving in humans.
December 17, 2025 at 3:06 PM
As long as the animal reservoir is a well-mixed population that can be modeled by the Poisson distribution from which D=2 is drawn, this seems to be correct by my understanding. I don't know if that is biologically a good model though as opposed to an I1 evolving in humans.
No problem. The D=2 dilution penalty is based on a reservoir clock modeling that Pekar makes no statements about. That cannot be called 'Pekar's model'.
December 15, 2025 at 9:02 PM
No problem. The D=2 dilution penalty is based on a reservoir clock modeling that Pekar makes no statements about. That cannot be called 'Pekar's model'.
I1 has 'fiddly' parameters too e.g. stochasticity for Lineage A but not B to explain clock reversal although both A and B were pre-registered to be linked to the shared market prior to discovery; lack of intermediates in human evolution and using the debunked Lv et al as evidence for intermediates.
December 15, 2025 at 6:28 PM
I1 has 'fiddly' parameters too e.g. stochasticity for Lineage A but not B to explain clock reversal although both A and B were pre-registered to be linked to the shared market prior to discovery; lack of intermediates in human evolution and using the debunked Lv et al as evidence for intermediates.
Maybe..this is above my pay grade. But a median 8 spillovers (as Pekar suggested) over a period of say 20-30 days with pre-existing reservoir diversity may mean high probability that A and B spill over within 4 days of each other into a correlated introduction environment.
December 15, 2025 at 4:37 PM
Maybe..this is above my pay grade. But a median 8 spillovers (as Pekar suggested) over a period of say 20-30 days with pre-existing reservoir diversity may mean high probability that A and B spill over within 4 days of each other into a correlated introduction environment.
'Circular' 'phylogenetic data don't say anything' 'distractor seem too strong when the complaint is about prior-sensitivity. Do you agree though that if there were multiple introductions, your analysis for exactly two introductions does not capture that?
December 15, 2025 at 2:39 AM
'Circular' 'phylogenetic data don't say anything' 'distractor seem too strong when the complaint is about prior-sensitivity. Do you agree though that if there were multiple introductions, your analysis for exactly two introductions does not capture that?
The way I understood this paper, B is 99% monophyletic, 69% >=1 B spillovers, 82.6% are multiple spillovers which then (to me) meant single-spillovers are in A and 69% of trees involve both A and B. I could be mistaken. www.biorxiv.org/content/10.1...
www.biorxiv.org
December 14, 2025 at 8:02 PM
The way I understood this paper, B is 99% monophyletic, 69% >=1 B spillovers, 82.6% are multiple spillovers which then (to me) meant single-spillovers are in A and 69% of trees involve both A and B. I could be mistaken. www.biorxiv.org/content/10.1...
I'm referring to Hensel/deBarre '25
December 14, 2025 at 7:46 PM
I'm referring to Hensel/deBarre '25
OK, I see what you are saying. If the goal is 'how Pekar is wrong', two intros in Pekar referred to separate A and B. If the goal is truth-seeking, any/only two intros including same lineage is not appropriate as phylodynamic analysis suggests 2 separate lineage introduction as the most likely tree.
December 14, 2025 at 7:15 PM
OK, I see what you are saying. If the goal is 'how Pekar is wrong', two intros in Pekar referred to separate A and B. If the goal is truth-seeking, any/only two intros including same lineage is not appropriate as phylodynamic analysis suggests 2 separate lineage introduction as the most likely tree.
Simply speaking, if you have pre-existing diversity in a market reservoir with spillovers linked to a short window in the market, near-simultaneous introductions are the norm, and there isn't much of a dilution factor.
December 12, 2025 at 12:17 AM
Simply speaking, if you have pre-existing diversity in a market reservoir with spillovers linked to a short window in the market, near-simultaneous introductions are the norm, and there isn't much of a dilution factor.
This would result in a probability mass that is high near where ta and tb are naturally close in a short window shared-market prior even before conditioning on the data where ta=tb is common, not rare.
December 12, 2025 at 12:11 AM
This would result in a probability mass that is high near where ta and tb are naturally close in a short window shared-market prior even before conditioning on the data where ta=tb is common, not rare.
It is about a shared-market prior with a short window where near simultaneous introductions would be common. Your calc appears to be a a maximum-likelihood evaluation of max(ta,tb) P(data|I2). The Bayes factor ought to be P(data|I2) - integ(P(data|ta,tb,I2) P(ta,tb|I2)dtadtb
December 12, 2025 at 12:10 AM
It is about a shared-market prior with a short window where near simultaneous introductions would be common. Your calc appears to be a a maximum-likelihood evaluation of max(ta,tb) P(data|I2). The Bayes factor ought to be P(data|I2) - integ(P(data|ta,tb,I2) P(ta,tb|I2)dtadtb
I believe you are focusing on the most favorable likelihood point in parameter space of ta, tb for I2 but not on how much total probability mass is assigned to that region. It is not about whether the diagonal ta=tb can happen, but that this is a sliver of possibilities in (ta,tb) parameter space.
December 12, 2025 at 12:03 AM
I believe you are focusing on the most favorable likelihood point in parameter space of ta, tb for I2 but not on how much total probability mass is assigned to that region. It is not about whether the diagonal ta=tb can happen, but that this is a sliver of possibilities in (ta,tb) parameter space.
Plus the hamster example shows that multiple NT differences can easily be held in an animal reservoir in a shared-market environment.
December 10, 2025 at 8:06 PM
Plus the hamster example shows that multiple NT differences can easily be held in an animal reservoir in a shared-market environment.