sebjenseb

sebjenseb

Discussion about this post

User's avatar
Ino's avatar
Ino
Apr 15, 2024

It would be great to see estimation for Einstein and John von Neumann

Expand full comment
Reply
Share
PigeonPrincipality's avatar
PigeonPrincipality
Sep 30, 2023

My impression is that using Hoeflin test data in a regression model will always prove quite fruitless.

It's patently absurd to assume that the Mega Test, for example, matches the general population distribution of intelligence. Any claim on behalf of Omni that its readers have an average IQ of 140 SD16 is immaterial (and a distortion of evidence). Omni's total circulation substantially exceeded 700000 copies per month, of which only a miniscule and self-selected fraction bothered with sending in answers for these tests. What that number actually represents is data reported to Ronald Hoeflin regarding prior scores of those who did. There are of course plentiful reasons to doubt the validity of the official norm. Redvaldsen reanalyzed this data and concluded that officially issued scores were inflated by around ten points at the highest level. This study is unfortunately not of much higher quality than Hoeflin's own original data -- it downplays the value of the old SAT as a high-range test while instead regressing against antiquated editions of the SB. It is better to use the SAT in norming the Mega, since both are primarily crystallized (as suggested by the low correlations of the Mega with fluid tests). Another issue besides norm inflation comprises the low range-corrected correlations obtained from Dr. Hoeflin's tests against reliable and valid instruments. Correlations to tests such as the WAIS, SB, and CTMM average ~0.3 without correcting for range restriction, but not much higher after doing so.

Untimed hobby tests do not belong in the same regression model as properly standardized ones. I've taken a few. If I were to believe Hoeflin's Power Test, LS60, and SLSE 48, I would be forced to conclude my spatial ability is at least five standard deviations above the mean. This is exceedingly unlikely. Marilyn vos Savant's childhood score was improperly computed. On a deviation scale, I believe I recall the actual value (which her psychologist flubbed the arithmetic on even for ratio IQ) would be about 2 2/3 SDs.

Expand full comment
Reply
Share
4 more comments...

No posts

© 2025 Sebastian Jensen
Privacy ∙ Terms ∙ Collection notice
Start writingGet the app
Substack is the home for great culture