3 ms·

So a psychologist who wrote a book condemning the original twins IQ study now publishes a paper where she purportedly refutes that study's most famous findings

So a psychologist who wrote a book condemning the original twins IQ study now publishes a paper where she purportedly refutes that study's most famous findings based on the analysis of a single set of separated twins? Isn't this the ultimate in p-hacking? Is there any reason to draw any conclusions from a sample size of one?

michaelmcmillan 4y ago
Exactly, this is ridiculous...
VictorPath 4y ago
An unasked question is how many of the identical twins in the original study were separated at birth and raised in very different environments? The answer is not many. A large number of the separated twins were separated at an older age, or were adopted by a close family member of a similar class background. I don't see this study as any more muddled than the original study.
- ravi-delia 4y ago
  To be clear, one of these twins suffered several concussions while the other did not
- car_analogy 4y ago
  Other twin studies didn't look solely at twins, but compared them with non-twins. As poster civilized said https://news.ycombinator.com/item?id=31304805 https://news.ycombinator.com/item?id=31304805: monozygotic twins reared apart show much more similar IQs than unrelated individuals raised together, indicating that genetics are the dominant driver of IQ. https://www.gwern.net/docs/iq/1996-bouchard-2.pdf https://www.gwern.net/docs/iq/1996-bouchard-2.pdf
littlestymaar 4y ago
Increasing the sample size is useful to get nice statistical properties out of noisy data. That doesn't mean you cannot learn anything from tiny samples, especially when the data isn't noisy (which is the case here, since they have the same set of genes, a lot of the noise is gone).
itsdrewmiller 4y ago
This is not an example of p-hacking. P-hacking is when you test so many different hypotheses that some of them randomly happen to have low p-values.

michaelmcmillan 4y ago

Exactly, this is ridiculous...

VictorPath 4y ago

An unasked question is how many of the identical twins in the original study were separated at birth and raised in very different environments? The answer is not many. A large number of the separated twins were separated at an older age, or were adopted by a close family member of a similar class background. I don't see this study as any more muddled than the original study.

ravi-delia 4y ago
To be clear, one of these twins suffered several concussions while the other did not
car_analogy 4y ago
Other twin studies didn't look solely at twins, but compared them with non-twins. As poster civilized said https://news.ycombinator.com/item?id=31304805 https://news.ycombinator.com/item?id=31304805: monozygotic twins reared apart show much more similar IQs than unrelated individuals raised together, indicating that genetics are the dominant driver of IQ. https://www.gwern.net/docs/iq/1996-bouchard-2.pdf https://www.gwern.net/docs/iq/1996-bouchard-2.pdf

littlestymaar 4y ago

Increasing the sample size is useful to get nice statistical properties out of noisy data. That doesn't mean you cannot learn anything from tiny samples, especially when the data isn't noisy (which is the case here, since they have the same set of genes, a lot of the noise is gone).

itsdrewmiller 4y ago

This is not an example of p-hacking. P-hacking is when you test so many different hypotheses that some of them randomly happen to have low p-values.