Michael Weissman
@mbweissman.bsky.social
Yep, that rate is similar, with huge error bars due to low N. Just as expected if SC2 came in via the mammal it's most associated with, humans.
December 12, 2025 at 6:59 PM
Yep, that rate is similar, with huge error bars due to low N. Just as expected if SC2 came in via the mammal it's most associated with, humans.
I use precisely simultaneous spills. Read!
December 12, 2025 at 1:00 AM
I use precisely simultaneous spills. Read!
I use precisely ta-tb -0 with no probabalistic range to maximize 2-spill odds. I allow a small range for ta +tb, although even exact fine-tuning hardly helps 2-spill. This is all spelled out in simple words and equations.
www.science.org/doi/10.1126/...
arxiv.org/abs/2510.01484
www.science.org/doi/10.1126/...
arxiv.org/abs/2510.01484
December 12, 2025 at 12:59 AM
I use precisely ta-tb -0 with no probabalistic range to maximize 2-spill odds. I allow a small range for ta +tb, although even exact fine-tuning hardly helps 2-spill. This is all spelled out in simple words and equations.
www.science.org/doi/10.1126/...
arxiv.org/abs/2510.01484
www.science.org/doi/10.1126/...
arxiv.org/abs/2510.01484
You may remember Galileo's words from the course you took here:
December 11, 2025 at 8:50 PM
You may remember Galileo's words from the course you took here:
good reason since it wasn't a big deal in the early stages.
You're conceding that a key paper, 29 authors, repeatedly referenced by you et al., hyped to NYT as "dispositive", missed all the ingredients needed to come to any conclusion. Because when its errors are fixed you don't like the answer.
You're conceding that a key paper, 29 authors, repeatedly referenced by you et al., hyped to NYT as "dispositive", missed all the ingredients needed to come to any conclusion. Because when its errors are fixed you don't like the answer.
December 11, 2025 at 8:09 PM
good reason since it wasn't a big deal in the early stages.
You're conceding that a key paper, 29 authors, repeatedly referenced by you et al., hyped to NYT as "dispositive", missed all the ingredients needed to come to any conclusion. Because when its errors are fixed you don't like the answer.
You're conceding that a key paper, 29 authors, repeatedly referenced by you et al., hyped to NYT as "dispositive", missed all the ingredients needed to come to any conclusion. Because when its errors are fixed you don't like the answer.
Ok, here's the answers>
- P(D=2) depends only on the expectation of reservoir diversity. I follow Bayes rules to tune that to favor I2.
- Unsuccessful spillovers are already allowed in the Pekar simulations, which I used without modification.
- immunological feedback was ignored by Pekar, with...
- P(D=2) depends only on the expectation of reservoir diversity. I follow Bayes rules to tune that to favor I2.
- Unsuccessful spillovers are already allowed in the Pekar simulations, which I used without modification.
- immunological feedback was ignored by Pekar, with...
December 11, 2025 at 8:05 PM
Ok, here's the answers>
- P(D=2) depends only on the expectation of reservoir diversity. I follow Bayes rules to tune that to favor I2.
- Unsuccessful spillovers are already allowed in the Pekar simulations, which I used without modification.
- immunological feedback was ignored by Pekar, with...
- P(D=2) depends only on the expectation of reservoir diversity. I follow Bayes rules to tune that to favor I2.
- Unsuccessful spillovers are already allowed in the Pekar simulations, which I used without modification.
- immunological feedback was ignored by Pekar, with...
True but this exchange has given me a sense of why he might have started to sound that way.
December 11, 2025 at 8:01 PM
True but this exchange has given me a sense of why he might have started to sound that way.
Them's a lot of words, none quite logical. Have you been replaced by an LLM?
December 11, 2025 at 5:01 PM
Them's a lot of words, none quite logical. Have you been replaced by an LLM?
The actual (incorrect) equations they used described in their Supplement are not mere "shorthand". Verbal blather does not replace math.
December 11, 2025 at 4:52 PM
The actual (incorrect) equations they used described in their Supplement are not mere "shorthand". Verbal blather does not replace math.
P2022 says "We assume each introduction is independent, allowing us to generalize this probability to P(𝜏|I_n)." Their polytomy probabilities mathematically assume that multiple intros are independent and of different sequences. Your epicycles are explicitly omitted from their math by them.
December 11, 2025 at 3:01 PM
P2022 says "We assume each introduction is independent, allowing us to generalize this probability to P(𝜏|I_n)." Their polytomy probabilities mathematically assume that multiple intros are independent and of different sequences. Your epicycles are explicitly omitted from their math by them.
Quick check : adding a 3d successful spill that happens to look just like A or B raises P(size ratio ok) by ~x1.3. The cost for sequence match (given post-intro to clade root differences) is x 0.25 or worse. You're barking up the wrong tree.
December 11, 2025 at 2:40 PM
Quick check : adding a 3d successful spill that happens to look just like A or B raises P(size ratio ok) by ~x1.3. The cost for sequence match (given post-intro to clade root differences) is x 0.25 or worse. You're barking up the wrong tree.
I just use Pekar's own simulations of their model.
December 11, 2025 at 2:36 PM
I just use Pekar's own simulations of their model.
Whoops, I misread. For I1 P2022 shows that P is reduced a lot by the range of clade sizes, similar to what I show for I2. Also low P for the missing intermediate in the data. Stepping outside the model, T-dependent ascertainment says why that isn't so surprising.
DOI 10.1016/S2468-2667(20)30089-X
DOI 10.1016/S2468-2667(20)30089-X
December 11, 2025 at 12:32 PM
Whoops, I misread. For I1 P2022 shows that P is reduced a lot by the range of clade sizes, similar to what I show for I2. Also low P for the missing intermediate in the data. Stepping outside the model, T-dependent ascertainment says why that isn't so surprising.
DOI 10.1016/S2468-2667(20)30089-X
DOI 10.1016/S2468-2667(20)30089-X
You can add more adjustable parameters to a favored hypothesis to make it fit better. But then it has to pay by marginalizing over prior distributions of the extra parameters. Extra spills can slightly increase P(similar size) but at big cost of low P(no intermediate & D<3)
December 11, 2025 at 2:14 AM
You can add more adjustable parameters to a favored hypothesis to make it fit better. But then it has to pay by marginalizing over prior distributions of the extra parameters. Extra spills can slightly increase P(similar size) but at big cost of low P(no intermediate & D<3)
That sort of verbal BS is embarrassing even to read. Sounds like lines from The Simpsons.
December 11, 2025 at 1:58 AM
That sort of verbal BS is embarrassing even to read. Sounds like lines from The Simpsons.
The analysis already stratifies on successful, i.e. non-cryptic transmissions. I.e. "I2" already allows many spills but only 2 succeed. The P2022 description is clear that they mean 2 succeed. If you want more to succeed, you need to pay a likelihood for no intermediates, for D<3 etc.
December 11, 2025 at 1:55 AM
The analysis already stratifies on successful, i.e. non-cryptic transmissions. I.e. "I2" already allows many spills but only 2 succeed. The P2022 description is clear that they mean 2 succeed. If you want more to succeed, you need to pay a likelihood for no intermediates, for D<3 etc.
How many epicycles and equants would you like?
December 11, 2025 at 1:12 AM
How many epicycles and equants would you like?
I have no model. In Pekar 2022's own simulations of their own model of their own data only 16% of their sim pairs are close enough in size. Even for an unrealistic fine-tuned prior diversity, Poisson says at most 27% of the pairs have D=2. Slightly less fine-tuned, that drops to 20%. 0.2*0.16=0.03.
December 10, 2025 at 9:45 PM
I have no model. In Pekar 2022's own simulations of their own model of their own data only 16% of their sim pairs are close enough in size. Even for an unrealistic fine-tuned prior diversity, Poisson says at most 27% of the pairs have D=2. Slightly less fine-tuned, that drops to 20%. 0.2*0.16=0.03.
I already allow tuning ta-tb to 0 to give two-spill as good a shot as possible. You seem not to have even skimmed the elletr much less read the longer version.
December 10, 2025 at 9:39 PM
I already allow tuning ta-tb to 0 to give two-spill as good a shot as possible. You seem not to have even skimmed the elletr much less read the longer version.
Yep, already factored in.
December 10, 2025 at 9:38 PM
Yep, already factored in.
Seriously? I read those posts. That's what you're going with?
December 10, 2025 at 7:44 PM
Seriously? I read those posts. That's what you're going with?
I'd like to hear what you and your friends think about the simple basic math logic errors of Pekar 2022 before rehashing your subjective opinions on other questions.
December 10, 2025 at 2:47 PM
I'd like to hear what you and your friends think about the simple basic math logic errors of Pekar 2022 before rehashing your subjective opinions on other questions.
Any thoughts on the mathematical specifics of that eletter?
December 10, 2025 at 2:05 AM
Any thoughts on the mathematical specifics of that eletter?
TBC, Neil has blocked me so I don't know most of what is going on in this thread.
December 9, 2025 at 8:21 PM
TBC, Neil has blocked me so I don't know most of what is going on in this thread.