Fishing for correlations is a statistically dubious practice, but also fun. Some interesting ones (none were very high, except e.g. Father Age and Mother Age):
IQ and Hours Writing have correlation 0.26 (75 degrees), which is the only interesting IQ correlation.
Siblings and Older siblings have correlation 0.48 (61 degrees), which isn’t too surprising , but makes me wonder: do we expect this correlation to be 0.5 in general?
Most of the Big Five answers are slightly correlated (around +/-0.25, or 90+/-15 degrees) with each other, but not with anything else except the Autism Score. Shouldn’t well-designed personality traits be orthogonal, ideally?
CFAR question 7 (guess of height of redwood) was negatively correlated with Height (-0.23, or 103 degrees). No notable correlation with the random number, though.
CFAR question 7 (guess of height of redwood) was negatively correlated with Height (-0.23, or 103 degrees). No notable correlation with the random number, though.
I looked at this with the data set that I used for my CFAR analyses, and this correlation did not show up; r=-.02 (p=.58). On closer inspection, the correlation is present in the complete un-cleaned-up data set (r=-.21), but it is driven entirely by a single outlier who listed their own height as 13 cm and the height of the tallest redwood as 10,000 ft.
(In my analyses of the anchoring question I had excluded the data of 2 outliers who listed redwood heights of 10,000 ft. and 1 ft. Before running this correlation with Height, I checked the distribution and excluded everyone who listed a height under 100 cm, since those probably represent units confusions.)
Most of the Big Five answers are slightly correlated (around +/-0.25, or 90+/-15 degrees) with each other, but not with anything else except the Autism Score. Shouldn’t well-designed personality traits be orthogonal, ideally?
It might just pick out the cluster of “Less Wrong personality type”.
CFAR question 7 (guess of height of redwood) was negatively correlated with Height
Obviously it’s a matter of perspective. Tall people just tower over those redwoods.
It might just pick out the cluster of “Less Wrong personality type”.
In that case, it says something about the cluster as well. For example, Openness and Extraversion wouldn’t be positively correlated just because most LWers are both open and extraverted (or because most LWers are closed and introverted). We’d have to have something that specifically makes “open and extraverted” more likely to happen together than individually.
“Siblings and Older siblings have correlation 0.48 (61 degrees), which isn’t too surprising , but makes me wonder: do we expect this correlation to be 0.5 in general?”
In every sibling relationship, there is one older and one younger sibling, so half of all siblings are older siblings—a line with slope 0.5.
So apparently, according to tradition the twin that is conceived first is believed to be born last, and there are folk explanations like “The first conceived attaches to the uterus first, so is more firmly stuck”. And so the later-born is considered oldest due to having been conceived first, even though that is not even a thing that can happen in the case of identical twins.
Still tracking down a decent history of the phenomenon, but it’s an interesting start.
It’s something I heard from my uncles (a pair of twins) and their mother. I can find stuff online, but it’s in Italian. Googling for gemello più vecchio (Italian for ‘older twin’) does turn up relevant stuff, so it’s not something my grandma made up. EDIT: apparently there was a myth about the first to be conceived is the last to be born (which for identical twins is Not Even Wrong). Someone answered on Google Answers, “if you went in a phone booth with a friend, the first of you to get in would be the last to come out, wouldn’t she?”
Fishing for correlations is a statistically dubious practice, but also fun. Some interesting ones (none were very high, except e.g. Father Age and Mother Age):
IQ and Hours Writing have correlation 0.26 (75 degrees), which is the only interesting IQ correlation.
Siblings and Older siblings have correlation 0.48 (61 degrees), which isn’t too surprising , but makes me wonder: do we expect this correlation to be 0.5 in general?
Most of the Big Five answers are slightly correlated (around +/-0.25, or 90+/-15 degrees) with each other, but not with anything else except the Autism Score. Shouldn’t well-designed personality traits be orthogonal, ideally?
CFAR question 7 (guess of height of redwood) was negatively correlated with Height (-0.23, or 103 degrees). No notable correlation with the random number, though.
I looked at this with the data set that I used for my CFAR analyses, and this correlation did not show up; r=-.02 (p=.58). On closer inspection, the correlation is present in the complete un-cleaned-up data set (r=-.21), but it is driven entirely by a single outlier who listed their own height as 13 cm and the height of the tallest redwood as 10,000 ft.
(In my analyses of the anchoring question I had excluded the data of 2 outliers who listed redwood heights of 10,000 ft. and 1 ft. Before running this correlation with Height, I checked the distribution and excluded everyone who listed a height under 100 cm, since those probably represent units confusions.)
It might just pick out the cluster of “Less Wrong personality type”.
Obviously it’s a matter of perspective. Tall people just tower over those redwoods.
In that case, it says something about the cluster as well. For example, Openness and Extraversion wouldn’t be positively correlated just because most LWers are both open and extraverted (or because most LWers are closed and introverted). We’d have to have something that specifically makes “open and extraverted” more likely to happen together than individually.
Something like Berkson’s paradox (people who are neither open nor introverted are unlikely to read LW)?
Good point. Objection retracted (in the conversational sense).
“Siblings and Older siblings have correlation 0.48 (61 degrees), which isn’t too surprising , but makes me wonder: do we expect this correlation to be 0.5 in general?”
In every sibling relationship, there is one older and one younger sibling, so half of all siblings are older siblings—a line with slope 0.5.
A correlation coefficient is not a slope. (A slope changes if you multiply one of the variables by a constant, whereas a correlation doesn’t.)
EDIT: I think the slope is the correlation times the standard deviation of y divided by the standard deviation of x.
How would twins reply to these questions?
All the twins I’ve known have regarded the first-born as “older” (and one has been first-born).
In Italy traditionally it’s the other way round. (Don’t ask me why.)
So apparently, according to tradition the twin that is conceived first is believed to be born last, and there are folk explanations like “The first conceived attaches to the uterus first, so is more firmly stuck”. And so the later-born is considered oldest due to having been conceived first, even though that is not even a thing that can happen in the case of identical twins.
Still tracking down a decent history of the phenomenon, but it’s an interesting start.
That sounds very counterintuitive. Do you have a citation? I can’t find information online.
It’s something I heard from my uncles (a pair of twins) and their mother. I can find stuff online, but it’s in Italian. Googling for gemello più vecchio (Italian for ‘older twin’) does turn up relevant stuff, so it’s not something my grandma made up. EDIT: apparently there was a myth about the first to be conceived is the last to be born (which for identical twins is Not Even Wrong). Someone answered on Google Answers, “if you went in a phone booth with a friend, the first of you to get in would be the last to come out, wouldn’t she?”
My Italian should be good enough for that. Grazie!