Under a simple assortative mating model written recently by Hays Golden (via MR), conservative* estimates of prevalence change will continue for multiple generations; excerpts.
* Conservative because his model assumes no underlying allele change, purely changes in mating patterns; in reality, we can expect increase in autism-related alleles, I think.
I’ve never quite bought the diagnosis thesis as a complete explanation—assortative mating seems plausible as a contributing factor, and implies the rate won’t stabilize anytime soon.
So I think this calls for a bet or at least more precise prediction: what probability do you assign to you being right; what exactly what rigorous objective data-source would you or a neutral third-party use to judge this; and what time-frame will this be true or false in?
(I would also like you to claim this account under either your real name or a pseudonym you’ve invested a lot in, so you can’t simply pretend it never happened if you’re wrong or lose the bet—give a reason to not ignore you as either willfully increasing counterparty risk or gambling on being right and then claiming it publicly in the future.)
Tumblr doesn’t let you respond to posts, so I’ll do it here:
You discuss a comic in a recent Tumblr post. The scene with the snowglobe is actually a reference I think you missed. Specifically, the last episode of the 1980s-era TV show St. Elsewhere, in which it’s suggested that the entire show took place in the imagination of a child whose parents say he is autistic.
From Wikipedia:
“The Last One”
The 1988 final episode of St. Elsewhere, known as “The Last One”, ended in a context very different from every other episode of the series. As the camera pans away from the snow beginning to fall at St. Eligius hospital, the scene changes to Donald Westphall’s autistic son Tommy, along with Daniel Auschlander in an apartment building. Westphall arrives home from a day’s work, and wears clothes suggesting that he is a construction worker. “Auschlander” is revealed to be Donald’s father, and thus Tommy’s grandfather. Donald laments to his father, “I don’t understand this autism. I talk to my boy, but...I’m not even sure if he ever hears me...Tommy’s locked inside his own world. Staring at that toy all day long. What does he think about?” The toy is revealed to be a snow globe with a replica of St. Eligius hospital inside. Tommy shakes the snow globe, and is told by his father to come and wash his hands, after having left the snow globe on the family’s television set.[1]
One of the more common interpretations[citation needed] of this scene is that as Tommy shakes the snow globe in the apartment, he also makes it snow at the “fictional” St. Eligius. His father and grandfather also seem to work at this hospital even though neither man has ever experienced such a role. By implication this interpretation suggests the total series of events in the series St. Elsewhere had been a product of Tommy Westphall’s imagination.
This comic is applying the same conceit to Foster’s Home For Imaginary Friends.
Thank you, I’ve now had time to read your post a second time and parse your other questions. Here are my answers.
(What is counterparty risk, by the way?)
The CDC will be an adequate source for this. I don’t know and can’t figure out how to find out when they’ll next release an estimate of how common ASDs are. If they do it before the DSM-V comes out or shortly after it comes out, then I would wait and take not the next but the one after. If they wait until the DSM-V has been in use for a while, and I’m not sure how long a while should be, then maybe just the next one.
I am not well-calibrated enough to give good probability estimates and I worry that I’d just be making something up if I tried to give a number. I also revised down slightly from six hours ago considering what you and Yvain have said. I still think having this out there, even imperfectly, is of social utility because until recently, the better diagnosis model and the epidemic model had both predicted increasing prevalence. I wouldn’t bet on it because, among other reasons, this prediction isn’t really about proving me right or wrong. It’s about proving a model right or wrong. If that model is wrong, then I update. I change my beliefs and actions, but then I move on. I don’t “change sides” or anything like that. However, the model stands or falls here (or at some future predetermined point farther on, if one in 86 is still not the correct prevalence) and if it’s going to survive, I want to see it believed by everyone, partly because proving this model eliminates the vaccine injury model (among others). You, who have reasoned that accumulated mutations are a reasonable model, are really not my opponent here.
I also admit that I want this model to be true because even without the hard work of educating people, it would show that there’s no epidemic to be scared of, which might make educating people easier. That could cloud my judgment, but an empirical test will not have clouded judgment.
The CDC will be an adequate source for this. I don’t know and can’t figure out how to find out when they’ll next release an estimate of how common ASDs are.
My general experience with this sort of thing is that the dataset will be rich and include many different ways of slicing the data and with multiple definitions, so it’s possible to pick and choose numbers to get what you want. Hence, specifying ‘CDC’ is better than specifying no source of data at all, but to make a good prediction one should specify the specific metric in the CDC surveys or reports.
I am not well-calibrated enough to give good probability estimates and I worry that I’d just be making something up if I tried to give a number. I also revised down slightly from six hours ago considering what you and Yvain have said.
If you can’t give a good probability estimate, then no one here will take you seriously nor will you be able to make or accept bets (so no one outside here will take you seriously either). If you want to learn how to do this, my own opinions can be found in http://www.gwern.net/Prediction%20markets
You, who have reasoned that accumulated mutations are a reasonable model, are really not my opponent here.
That’s Yvain, not me. I read the recent correlative study like everyone else and find it interesting, but I’m not sure how persuasive I find it.
Fair enough?
If you’re really serious about it… What I would’ve done, if I had this sort of big public statement in mind, is written up a brief precis of why the diagnosis model is right, why I expect diagnosis-driven increases to level off soon (in stock market shorting, they have an expression: ‘don’t try to catch a falling knife’. When you short a stock, knowing when to start shorting is as important as knowing what is overpriced, because ‘the market can stay irrational longer than you can stay solvent’), what my exact metrics will be, what probability I assign over what periods, cryptographically sign my statement, and challenge some prominent anti-diagnosticers to bets on those terms—ideally through LongBets.org for nice round sums like $100, $500, or $1000.
Is this a lot of work? Yes! But you are trying to buy credibility, attention, and change minds down the line; you should not expect that to be cheap or easy.
Especially since he immediately stopped commenting (last comment November 2012), and there’s no way to track him down to a real name, so if one had made a bet with him and won, how could one collect?
Under a simple assortative mating model written recently by Hays Golden (via MR), conservative* estimates of prevalence change will continue for multiple generations; excerpts.
* Conservative because his model assumes no underlying allele change, purely changes in mating patterns; in reality, we can expect increase in autism-related alleles, I think.
I’ve never quite bought the diagnosis thesis as a complete explanation—assortative mating seems plausible as a contributing factor, and implies the rate won’t stabilize anytime soon.
So I think this calls for a bet or at least more precise prediction: what probability do you assign to you being right; what exactly what rigorous objective data-source would you or a neutral third-party use to judge this; and what time-frame will this be true or false in?
(I would also like you to claim this account under either your real name or a pseudonym you’ve invested a lot in, so you can’t simply pretend it never happened if you’re wrong or lose the bet—give a reason to not ignore you as either willfully increasing counterparty risk or gambling on being right and then claiming it publicly in the future.)
I hereby claim that I am not-allistic on tumblr.
Tumblr doesn’t let you respond to posts, so I’ll do it here:
You discuss a comic in a recent Tumblr post. The scene with the snowglobe is actually a reference I think you missed. Specifically, the last episode of the 1980s-era TV show St. Elsewhere, in which it’s suggested that the entire show took place in the imagination of a child whose parents say he is autistic.
From Wikipedia:
This comic is applying the same conceit to Foster’s Home For Imaginary Friends.
Then St. Elsewhere, rather than the comic, is problematic and wrong. Thank you; I wouldn’t have known that had you not told me.
I’m still glad to have put that up on tumblr because I still don’t want people thinking that’s an accurate portrayal of autism.
I have no idea who that is, and googling I just see a bunch of tags. (Also, you didn’t answer the other questions.)
Tumblr urls are formed like this:
http://not-allistic.tumblr.com/
Thank you, I’ve now had time to read your post a second time and parse your other questions. Here are my answers.
(What is counterparty risk, by the way?)
The CDC will be an adequate source for this. I don’t know and can’t figure out how to find out when they’ll next release an estimate of how common ASDs are. If they do it before the DSM-V comes out or shortly after it comes out, then I would wait and take not the next but the one after. If they wait until the DSM-V has been in use for a while, and I’m not sure how long a while should be, then maybe just the next one.
I am not well-calibrated enough to give good probability estimates and I worry that I’d just be making something up if I tried to give a number. I also revised down slightly from six hours ago considering what you and Yvain have said. I still think having this out there, even imperfectly, is of social utility because until recently, the better diagnosis model and the epidemic model had both predicted increasing prevalence. I wouldn’t bet on it because, among other reasons, this prediction isn’t really about proving me right or wrong. It’s about proving a model right or wrong. If that model is wrong, then I update. I change my beliefs and actions, but then I move on. I don’t “change sides” or anything like that. However, the model stands or falls here (or at some future predetermined point farther on, if one in 86 is still not the correct prevalence) and if it’s going to survive, I want to see it believed by everyone, partly because proving this model eliminates the vaccine injury model (among others). You, who have reasoned that accumulated mutations are a reasonable model, are really not my opponent here.
I also admit that I want this model to be true because even without the hard work of educating people, it would show that there’s no epidemic to be scared of, which might make educating people easier. That could cloud my judgment, but an empirical test will not have clouded judgment.
Fair enough?
My general experience with this sort of thing is that the dataset will be rich and include many different ways of slicing the data and with multiple definitions, so it’s possible to pick and choose numbers to get what you want. Hence, specifying ‘CDC’ is better than specifying no source of data at all, but to make a good prediction one should specify the specific metric in the CDC surveys or reports.
If you can’t give a good probability estimate, then no one here will take you seriously nor will you be able to make or accept bets (so no one outside here will take you seriously either). If you want to learn how to do this, my own opinions can be found in http://www.gwern.net/Prediction%20markets
That’s Yvain, not me. I read the recent correlative study like everyone else and find it interesting, but I’m not sure how persuasive I find it.
If you’re really serious about it… What I would’ve done, if I had this sort of big public statement in mind, is written up a brief precis of why the diagnosis model is right, why I expect diagnosis-driven increases to level off soon (in stock market shorting, they have an expression: ‘don’t try to catch a falling knife’. When you short a stock, knowing when to start shorting is as important as knowing what is overpriced, because ‘the market can stay irrational longer than you can stay solvent’), what my exact metrics will be, what probability I assign over what periods, cryptographically sign my statement, and challenge some prominent anti-diagnosticers to bets on those terms—ideally through LongBets.org for nice round sums like $100, $500, or $1000.
Is this a lot of work? Yes! But you are trying to buy credibility, attention, and change minds down the line; you should not expect that to be cheap or easy.
The risk that if you lose the bet, you don’t pay up. It’s relatively high here because a single-purpose account has no reputation to care about.
Especially since he immediately stopped commenting (last comment November 2012), and there’s no way to track him down to a real name, so if one had made a bet with him and won, how could one collect?