I dunno, this is probably the best-case for IQ, to be honest.
Remember kids, IQ was developed to predict which children would perform best in school so that they could be streamed, and that's what it's best at.
I don't think that the author is abusing statistics here at all, as the variance explained by IQ on grades is always pretty large (in psychology, 35% explained is super, super good), and if I really cared (i.e. you reply and ask for this) I could probably dig up another few datasets that show the same or higher association.
"(in 1904) The French Ministry of Education asked these researchers to develop a test that would allow for distinguishing mentally retarded children from normally intelligent, but lazy children." -- Binet developed it out of an earlier concept of 'mental age', the average ability level of a child aged so many years and months.
I'm an engineer and if I present any model with an R2 of 0.35 it would be flat out rejected because it would be worse than random chance and they won't use the model to make decisions.
Assuming the R2 here is indeed correct how do you use this to make real life hiring or admission decisions ? Just weight it at 35% ? What do you do for the other 65% then ?
> if I present any model with an R2 of 0.35 it would be flat out rejected because it would be worse than random chance
If I was interviewing an engineer that said something like this, I wouldn't hire them. It's somewhat unclear what you even think "random chance" means in a regression context.
I was simply agreeing with the parent that the standard for evidence in the social sciences is a lot weaker than would be acceptable in physics, chemistry, etc.
Necessarily weaker, though. You can't break out an electron microscope and numerically confirm observations about emergent outcomes from a billion individuals with their own psychologies. Either we allow for "softer" observations/experiments or we throw up our hands and give up.
You are welcome to do so if you have the time. To be honest though, I'm not sure what that would achieve. I don't think that these kinds of claims can be seriously supported by a discussion in the format of an online message board, no matter how many links and datasets we throw at each other.
Remember kids, IQ was developed to predict which children would perform best in school so that they could be streamed, and that's what it's best at.
I don't think that the author is abusing statistics here at all, as the variance explained by IQ on grades is always pretty large (in psychology, 35% explained is super, super good), and if I really cared (i.e. you reply and ask for this) I could probably dig up another few datasets that show the same or higher association.