5 ms·

I can't understand your response at all. The article shows that a flame-retardent increases risk of autoimmune condition with 94% confidence. And you downplay

I can't understand your response at all. The article shows that a flame-retardent increases risk of autoimmune condition with 94% confidence. And you downplay it because... it's less than 95% confident? (by the way, the paper finds its main finding at p < .02) To my mind, a major post-mortem is necessary, government funded. How many other flame-retardents cause health issues? How did they get past approval? In the meantime, an immediate injunction against flame-retardent laws should be passed (numerous tobacco-lobbied laws REQUIRE retardents), and all used retardents should be listed on the labels of all products. I expect the people in charge of public health to be as anal about their job as I am about mine.

JoeAltmaier 6y ago
94% is an ordinary 'coincidence' value. Flip a coin until it comes up heads 5 times in a row. What are the odds! Well, if you keep flipping you always can see that. It's inevitable. That's why investigation is done until there are many more 9's. Because one 9 can be found just by trying a bunch of stuff.
- zug_zug 6y ago
  So the statistical term for what you're saying is that the paper didn't use bonferroni correction. Which is relevant (but unrelated to the p < .05 arbitrary boundary). > That's why investigation is done until there are many more 9's Nobody said the investigation is done. What is being said is that this alone is a red flag that warrants a lot of investigation (and the investigation shouldn't have to be research by colleges whenever a graduate student gets around to it). This is something the EPA should fund an attempt to reproduce. It's the EPA's job to prevent dangerous chemicals from being sold in public goods. It's not college's job to need to conduct exhausting studies with absolute proof with many 9s that something is unsafe and then hand it off to the EPA to act. Once any reasonable doubt is raised the EPA needs to take ownership.
  - JoeAltmaier 6y ago
    This investigation is the bare beginning. It may turn out to be completely wrong. It does not warrant 'a lot of investigation'. There are so many spurious results out there, it would be a fools errand to dig into every curious result. And no I did not mean some correction. I meant, that publishing is a filter for any seemingly correlated result. It will inevitably include simple coincidences, because of the sample sizes involved. To put it simply, all coincidental results end up being published, because they look so remarkable. It's a reason to be very skeptical of small samples.
    - zug_zug 6y ago
      >>> I meant, that publishing is a filter for any seemingly correlated result. Great, we all know what publication bias is. Sure there's a non-zero chance that this is a statistical anomaly. But my dude, what exactly is going on in your head? Numerous flame retardants have already been found dangerous and toxic [1]. Are you seriously concerned people are going to take their health too seriously and throw away their couches because of this study? I think the exponentially more likely, and exponentially more dangerous concern is these flame-retardant chemicals aren't taken seriously enough... particularly since it's debatable whether they were ever necessary in the first place. 1 - https://en.wikipedia.org/wiki/Flame_retardant#Health_concerns https://en.wikipedia.org/wiki/Flame_retardant#Health_concern...
      - JoeAltmaier 6y ago
        What's going on is, the 'emotional outrage of the minute' eco-topics that keep coming up. I have limited investment to spend in outrage. I'm not spending it on every preliminary report that might show some correlation. For instance, tiny cats (6lbs? 10?) that sleep on furniture for 8-10 hours a day show elevated flame retardant in an absorbent tag. Do the cats respond in the same way as humans to these chemicals? (No) Is the acceptable dose for a 10lb cat different from a 100lb person? (Yes) Was anything else measured in the experiment (No?) Did the experiment seek out any other explanations? (No) This reeks of a cherry-picked goal-directed experiment to create outrage. It seems the opposite of science.
epmaybe 6y ago
Ah, I see what you're saying on the survival curve, thanks for the point out! I agree that's an interesting statistic, and it'll be interesting to see further studies investigating flame retardants. I'd be very interested in human studies, as I myself have thyroid issues (albeit more familial than anything). That being said, I'm not sure that you can claim that hyperthyroidism in cats is an autoimmune condition. I also am interested in what the effect size of this correlation actually is. We see the data that says that cats in 1995 had 20x lower rates of hyperthyroidism, but we don't have any additional information on how many cats were actually being screened for that. It's like saying (hopefully this isn't too controversial) the prevalence of autism is higher today than it was 20 years ago. That's an interesting question you bring up though, whether or not we should immediately intervene based on animal studies. And this is something we should probably extend to human studies, and as far low level as basic science studies. Where do you draw the line? Do you ban cell phones? Wifi? Milk? Meat? Coffee? Aspartame? And finally - Do you require all compounds to be listed on labels of all products? Or just some? How many compounds? Only when we find correlations? Or prior to that based on empirical evidence?

JoeAltmaier 6y ago

94% is an ordinary 'coincidence' value. Flip a coin until it comes up heads 5 times in a row. What are the odds! Well, if you keep flipping you always can see that. It's inevitable. That's why investigation is done until there are many more 9's. Because one 9 can be found just by trying a bunch of stuff.

zug_zug 6y ago
So the statistical term for what you're saying is that the paper didn't use bonferroni correction. Which is relevant (but unrelated to the p < .05 arbitrary boundary). > That's why investigation is done until there are many more 9's Nobody said the investigation is done. What is being said is that this alone is a red flag that warrants a lot of investigation (and the investigation shouldn't have to be research by colleges whenever a graduate student gets around to it). This is something the EPA should fund an attempt to reproduce. It's the EPA's job to prevent dangerous chemicals from being sold in public goods. It's not college's job to need to conduct exhausting studies with absolute proof with many 9s that something is unsafe and then hand it off to the EPA to act. Once any reasonable doubt is raised the EPA needs to take ownership.
- JoeAltmaier 6y ago
  This investigation is the bare beginning. It may turn out to be completely wrong. It does not warrant 'a lot of investigation'. There are so many spurious results out there, it would be a fools errand to dig into every curious result. And no I did not mean some correction. I meant, that publishing is a filter for any seemingly correlated result. It will inevitably include simple coincidences, because of the sample sizes involved. To put it simply, all coincidental results end up being published, because they look so remarkable. It's a reason to be very skeptical of small samples.
  - zug_zug 6y ago
    >>> I meant, that publishing is a filter for any seemingly correlated result. Great, we all know what publication bias is. Sure there's a non-zero chance that this is a statistical anomaly. But my dude, what exactly is going on in your head? Numerous flame retardants have already been found dangerous and toxic [1]. Are you seriously concerned people are going to take their health too seriously and throw away their couches because of this study? I think the exponentially more likely, and exponentially more dangerous concern is these flame-retardant chemicals aren't taken seriously enough... particularly since it's debatable whether they were ever necessary in the first place. 1 - https://en.wikipedia.org/wiki/Flame_retardant#Health_concerns https://en.wikipedia.org/wiki/Flame_retardant#Health_concern...
    - JoeAltmaier 6y ago
      What's going on is, the 'emotional outrage of the minute' eco-topics that keep coming up. I have limited investment to spend in outrage. I'm not spending it on every preliminary report that might show some correlation. For instance, tiny cats (6lbs? 10?) that sleep on furniture for 8-10 hours a day show elevated flame retardant in an absorbent tag. Do the cats respond in the same way as humans to these chemicals? (No) Is the acceptable dose for a 10lb cat different from a 100lb person? (Yes) Was anything else measured in the experiment (No?) Did the experiment seek out any other explanations? (No) This reeks of a cherry-picked goal-directed experiment to create outrage. It seems the opposite of science.

epmaybe 6y ago

Ah, I see what you're saying on the survival curve, thanks for the point out! I agree that's an interesting statistic, and it'll be interesting to see further studies investigating flame retardants. I'd be very interested in human studies, as I myself have thyroid issues (albeit more familial than anything). That being said, I'm not sure that you can claim that hyperthyroidism in cats is an autoimmune condition. I also am interested in what the effect size of this correlation actually is. We see the data that says that cats in 1995 had 20x lower rates of hyperthyroidism, but we don't have any additional information on how many cats were actually being screened for that. It's like saying (hopefully this isn't too controversial) the prevalence of autism is higher today than it was 20 years ago. That's an interesting question you bring up though, whether or not we should immediately intervene based on animal studies. And this is something we should probably extend to human studies, and as far low level as basic science studies. Where do you draw the line? Do you ban cell phones? Wifi? Milk? Meat? Coffee? Aspartame? And finally - Do you require all compounds to be listed on labels of all products? Or just some? How many compounds? Only when we find correlations? Or prior to that based on empirical evidence?