4 ms·
>> Variance explained by IQ The author performs a regression on the WLS dataset. I did not know what that is so I followed the links provided: >> The Wisconsi
by YeGoblynQueenne 6y ago
>> Variance explained by IQ
The author performs a regression on the WLS dataset. I did not know what that
is so I followed the links provided:
>> The Wisconsin Longitudinal Study (WLS) is a long-term study of a random sample
of 10,317 men and women who graduated from Wisconsin high schools in 1957.
So, in other words, it is a small dataset that is not representative of most
humans on the planet. It is also specifically populated by high school
graduates which is likely to introduce some unknown bias to the dataset- and
any results drawn from it in particular with regards to IQ.
In short, please don't abuse statistics, or if you do, don't do it in an
article that starts by accusing another text of being "just a rant".
- tgv 6y agoThe article quoted doesn't even mention the data or the test.
- disgruntledphd2 6y agoI 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.
- YeGoblynQueenne 6y agoYou 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.
- vsskanth 6y agoI'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 ?
- api 6y agoSocial sciences are called “soft” for a reason.
- vsskanth 6y agoWhy (asking seriously). I am not doubting the validity of the correlation. I just want to know how one can use this info properly.
- api 6y agoI 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.
- free_rms 6y agoNecessarily 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.
- whimsicalism 6y ago> it is a small dataset 10k is definitely large enough to draw conclusions. I think your concern about generalization is much more relevant.
- Terretta 6y agoBasing on 10,000 graduates of Wisconsin high schools in 1957 happens to “control for” (in a by-and-large way) many or most of the confounding influences one would have in a study today or in an area such as, for instance, Maryland or New York City. To your point about high school grads, though, it is absolutely worth noting that in 1957, “no child left behind” wasn’t a thing. In 1950, 36% of whites and 13% of blacks over age 25 had a high school diploma. In 2010, it was 92% and 84%. Then, graduation was generally a vocational choice by household and student rather than a performance bar. Today, it’s an externalized requirement, largely for the system to be funded: you will be graduated despite yourself. A few other diffs: 1954: Brown vs. Board of Education desegregation ordered 1951 spending per student: $215 for 155 days of school. 2010: $10,615 for 180 days of school. Average teacher salary then and now: $4K vs. $39K. 1955’s most popular youth book: Charlotte’s Web. 2012: Hunger Games.
- YeGoblynQueenne 6y ago>> Basing on 10,000 graduates of Wisconsin high schools in 1957 happens to “control for” (in a by-and-large way) many or most of the confounding influences one would have in a study today or in an area such as, for instance, Maryland or New York City. I don't understand how selecting a non-representative sub-population controls for anything! Same goes if the data was from Maryland or New York City. In any case, disentangling confounding factors and isolating the influence of IQ should be the goal of any research (though of course I can't ask that of an "analysis" published on a blog). If a dataset seems to reveal some deep relation between IQ and other outcomes because that dataset is free of confounding factors, that doesn't tell us what happens in the general case, where the confounding factors apply. In other words, there's a danger of finding what one is looking for because one is looking for it where it exists, when the goal of looking is to make sure it exists everywhere.