It is so huge that it is not possible. It is like saying that the children born in the pandemic are all 2-3 inches shorter than normal.
You know me as a big believer in the value of IQ in limited circumstances. I am not, however, a believer of childhood IQ's, and definitely not three year olds. Much too much volatility.
I agree with AVI that IQ tests on children aged 2 months to 3 years are probably bogus. What's more, I could make a more convincing case that sending most of the nation's children off to daycare tends to depress IQ than that staying home with Mom and Dad all day does.
Conveniently, Hanania, in his tweet thread, omitted the error bars on the developmental scores used for the comparison, and he omitted the error bars on the comparison's results.
Absent that context, there's no meaning to the IQ differences he claims the study found.
He[ck], even the inputs, from the "study" itself-- early learning composite (ELC), verbal development quotient (VDQ) and non-verbal development quotient (NVDQ) are entirely subjective measures, and there's no indication in the study of whether or how the assessors of those inputs were trained or normalized.
Separately, a word about stepwise regression. This is a statistics technique that keeps adding in additional factors (order often non-specific, even when order of appearance can itself make a difference--an order effect that linear regression (which this study's use of stepwise is), by its nature, is not equipped to detect) until the last add-in has no effect. That last one is then withdrawn and the analysis called complete. Stepwise regression can be very useful, but it's fraught with the potential for error and easily misused, just in doing the analysis.
ANCOVA on categorical data? Really? Analysis of Covariance, like its mate ANOVA (Analysis of Variance) assesses differences between/among means--which can't exist in categorical data. What's the average of some apples and some trees?
Enough statistics soapbox.
Back to my main point. The "study" does provide error ranges for its ANCOVA results, but those are meaningless as the ANCOVA was applied.
Might as well say the earlier cohort scored Bozos, the pandemic cohort ran laps, and the comparison results are magazines of interest.
6 comments:
It is so huge that it is not possible. It is like saying that the children born in the pandemic are all 2-3 inches shorter than normal.
You know me as a big believer in the value of IQ in limited circumstances. I am not, however, a believer of childhood IQ's, and definitely not three year olds. Much too much volatility.
I’m in no wise an expert on IQ, let alone in children; but environmental conditions can produce a height difference on that order.
https://www.bbc.com/news/magazine-17774210
I agree with AVI that IQ tests on children aged 2 months to 3 years are probably bogus. What's more, I could make a more convincing case that sending most of the nation's children off to daycare tends to depress IQ than that staying home with Mom and Dad all day does.
Conveniently, Hanania, in his tweet thread, omitted the error bars on the developmental scores used for the comparison, and he omitted the error bars on the comparison's results.
Absent that context, there's no meaning to the IQ differences he claims the study found.
He[ck], even the inputs, from the "study" itself-- early learning composite (ELC), verbal development quotient (VDQ) and non-verbal development quotient (NVDQ) are entirely subjective measures, and there's no indication in the study of whether or how the assessors of those inputs were trained or normalized.
Separately, a word about stepwise regression. This is a statistics technique that keeps adding in additional factors (order often non-specific, even when order of appearance can itself make a difference--an order effect that linear regression (which this study's use of stepwise is), by its nature, is not equipped to detect) until the last add-in has no effect. That last one is then withdrawn and the analysis called complete. Stepwise regression can be very useful, but it's fraught with the potential for error and easily misused, just in doing the analysis.
ANCOVA on categorical data? Really? Analysis of Covariance, like its mate ANOVA (Analysis of Variance) assesses differences between/among means--which can't exist in categorical data. What's the average of some apples and some trees?
Enough statistics soapbox.
Back to my main point. The "study" does provide error ranges for its ANCOVA results, but those are meaningless as the ANCOVA was applied.
Might as well say the earlier cohort scored Bozos, the pandemic cohort ran laps, and the comparison results are magazines of interest.
Eric Hines
You’re promoted to House Statistician. At least I think that’s a promotion.
Promoting me to House Liar, are you? [g]
Eric Hines
Post a Comment