US CDC plans study into vaccines and autism

[kj]

I know this is all over, but I have to get a word in.

"If the sample size is large enough, then all those factors you mentioned above will average out. The only thing that matters is the percentage of kids with Autism in vaccinated versus unvaccinated kids."

This is one of the most frustrating things you can say to a statistician. You have two groups of people, unvaccinated and vaccinated. You look at the number of people who have autism in each and compare them. You find a difference, perhaps 500 more people have autism in the vaccinated group. That's "more", but is it a statistically different amount? Iow, could you expect to see this difference occur just by chance? That's where statistics come in. You often see a confidence level like .005. If the difference is significant to the .005 level, it means you would expect to see it by chance only 5 times in 1000. So not very often, which means you can be confident it's not just by chance (it's a real difference).

You have to look at the size of the difference relative to how much "noise" or variability there is in the data. If it's really noisy, you need either a huge difference or a really big sample size to outweigh the noise. Confounds are part of what introduce noise, so you want to reduce those.

I think the hardest part of research like this is the actual "thing" that we call autism. The definition has changed continually throughout the years because we don't really know that much about it. Imagine if we were to find there are multiple causes for the symptoms we have decided identify as autism. In that case vaccines might affect the likelihood of one sub-type of autism, but not all the others, which would make the above comparison between vaccinated and unvaccinated groups less likely to show a difference. The effect would be "washed out". In other words, like Diana said, research is complex, and research involving humans is the most complex. It may seem simple, but it gets crazy real quick.

[Janit]

And here's another way to think about "correlation is not causation."

Let's speculate wildly:

Imagine that autism has a significant genetic component, and that this genetic component makes a person more vulnerable to other as yet unknown outside factors. Thus a child with identifiable autism will have parent(s) who carry the same gene(s) for autism susceptibility as well. The parent(s) may therefore themselves have mild undiagnosed "autistic" behaviors, one of which might be an affinity and comfort in rule-based decisions. One such rule-based decision might be to vaccinate their children.

This might result in a higher vaccination rate in children carrying the gene(s) for autism susceptibility. Thus the correct interpretation of any correlation between diagnosable autism and vaccines could be that autism causes vaccination rather than that vaccination causes autism.

Of course no such correlation has been reliably demonstrated, but even if it had been, there would still be more questions to ask to evaluate its meaning.

[kj]

Janit wrote:
And here's another way to think about "correlation is not causation."

Let's speculate wildly:

Imagine that autism has a significant genetic component, and that this genetic component makes a person more vulnerable to other as yet unknown outside factors. Thus a child with identifiable autism will have parent(s) who carry the same gene(s) for autism susceptibility as well. The parent(s) may therefore themselves have mild undiagnosed "autistic" behaviors, one of which might be an affinity and comfort in rule-based decisions. One such rule-based decision might be to vaccinate their children.

This might result in a higher vaccination rate in children carrying the gene(s) for autism susceptibility. Thus the correct interpretation of any correlation between diagnosable autism and vaccines could be that autism causes vaccination rather than that vaccination causes autism.

Of course no such correlation has been reliably demonstrated, but even if it had been, there would still be more questions to ask to evaluate its meaning.

(tu)

[Diana]

kj wrote:
[quote=Janit]
And here's another way to think about "correlation is not causation."

Let's speculate wildly:

Imagine that autism has a significant genetic component, and that this genetic component makes a person more vulnerable to other as yet unknown outside factors. Thus a child with identifiable autism will have parent(s) who carry the same gene(s) for autism susceptibility as well. The parent(s) may therefore themselves have mild undiagnosed "autistic" behaviors, one of which might be an affinity and comfort in rule-based decisions. One such rule-based decision might be to vaccinate their children.

This might result in a higher vaccination rate in children carrying the gene(s) for autism susceptibility. Thus the correct interpretation of any correlation between diagnosable autism and vaccines could be that autism causes vaccination rather than that vaccination causes autism.

Of course no such correlation has been reliably demonstrated, but even if it had been, there would still be more questions to ask to evaluate its meaning.

(tu)
Apparently there IS a genetic component.

[Janit]

Diana wrote:
[quote=kj]
[quote=Janit]
And here's another way to think about "correlation is not causation."

Let's speculate wildly:

Imagine that autism has a significant genetic component, and that this genetic component makes a person more vulnerable to other as yet unknown outside factors. Thus a child with identifiable autism will have parent(s) who carry the same gene(s) for autism susceptibility as well. The parent(s) may therefore themselves have mild undiagnosed "autistic" behaviors, one of which might be an affinity and comfort in rule-based decisions. One such rule-based decision might be to vaccinate their children.

This might result in a higher vaccination rate in children carrying the gene(s) for autism susceptibility. Thus the correct interpretation of any correlation between diagnosable autism and vaccines could be that autism causes vaccination rather than that vaccination causes autism.

Of course no such correlation has been reliably demonstrated, but even if it had been, there would still be more questions to ask to evaluate its meaning.

(tu)
Apparently there IS a genetic component.
Yes, I know. I phrased it this way to make the logical argument clear and to avoid going too far into the weeds of what "genetic" means. It was a rhetorical choice because I know that some people will go all glassy-eyed when the sciency language gets too dense.

[Diana]

Janit wrote:
[quote=Diana]
[quote=kj]
[quote=Janit]
And here's another way to think about "correlation is not causation."

Let's speculate wildly:

Imagine that autism has a significant genetic component, and that this genetic component makes a person more vulnerable to other as yet unknown outside factors. Thus a child with identifiable autism will have parent(s) who carry the same gene(s) for autism susceptibility as well. The parent(s) may therefore themselves have mild undiagnosed "autistic" behaviors, one of which might be an affinity and comfort in rule-based decisions. One such rule-based decision might be to vaccinate their children.

This might result in a higher vaccination rate in children carrying the gene(s) for autism susceptibility. Thus the correct interpretation of any correlation between diagnosable autism and vaccines could be that autism causes vaccination rather than that vaccination causes autism.

Of course no such correlation has been reliably demonstrated, but even if it had been, there would still be more questions to ask to evaluate its meaning.

(tu)
Apparently there IS a genetic component.
Yes, I know. I phrased it this way to make the logical argument clear and to avoid going too far into the weeds of what "genetic" means. It was a rhetorical choice because I know that some people will go all glassy-eyed when the sciency language gets too dense.
Ahh. Thanks. miley-laughing001:

[Janit]

Diana wrote:
[quote=Janit]
[quote=Diana]
[quote=kj]
[quote=Janit]
And here's another way to think about "correlation is not causation."

Let's speculate wildly:

Imagine that autism has a significant genetic component, and that this genetic component makes a person more vulnerable to other as yet unknown outside factors. Thus a child with identifiable autism will have parent(s) who carry the same gene(s) for autism susceptibility as well. The parent(s) may therefore themselves have mild undiagnosed "autistic" behaviors, one of which might be an affinity and comfort in rule-based decisions. One such rule-based decision might be to vaccinate their children.

This might result in a higher vaccination rate in children carrying the gene(s) for autism susceptibility. Thus the correct interpretation of any correlation between diagnosable autism and vaccines could be that autism causes vaccination rather than that vaccination causes autism.

Of course no such correlation has been reliably demonstrated, but even if it had been, there would still be more questions to ask to evaluate its meaning.

(tu)
Apparently there IS a genetic component.
Yes, I know. I phrased it this way to make the logical argument clear and to avoid going too far into the weeds of what "genetic" means. It was a rhetorical choice because I know that some people will go all glassy-eyed when the sciency language gets too dense.
Ahh. Thanks. miley-laughing001:
Rethinking this, I probably could have started with the observation that there IS a genetic component, and still kept it rhetorically simple. I trained as a molecular geneticist, and we have a habit of starting everything with the questions Why? and What if? It also serves me well for writing science fiction (educational :oldfogette, which is what I am doing now.