Because of the unusual nature of this test, this report will first consider its standard error of measurement. As the KIT can be taken multiple times, the standard error is directly observed in the standard deviation of the scores of repetitive KIT takers. This standard deviation is equal to the standard error of measurement.
For better understanding one must know the KIT has an internal method to estimate the standard error (SE) of each reported score, and in its initial period of use, until early March 2006, scores have consistently been reported with an estimated SE of 2.5 GT points or less. So .25 standard deviation on the high-range standard scale. From early March 2006 on, scores are being reported tighter, with the SE at 1.5 GT or less, to narrow the range of scores achieved by any individual. This is done to reduce the chance that someone taking the KIT many times will once "hit the jackpot" (get a score far above his real intelligence level).
Note that such an event - jackpot score - is theoretically possible on any IQ test, if one keeps taking test after test. But with the KIT, this risk is greater because one can take it many times. This report deals with KIT scores from the initial period with the estimated SE at 2.5. Here are the scores of the repetitive takers:
So the actual SE, computed as the weighted average of these standard deviations (SD), is 2.08 GT, or .208 SD on the high-range scale. That would be quite acceptable for a normal IQ test. But because the KIT can be taken many times, one would want it to be smaller. We may later see if the current convention of reporting scores with an estimated SE of 1.5 will indeed result in a smaller actual SE. To avoid confusion: a smaller SE does NOT mean LOWER scores on average, but a more narrow range of scores for any individual.
# testees: 13
Male mean: 53.62
Female mean: NaN (0 persons)
Note: there are a few mores scores that are not considered here because they are "or lower" ones of people whose real level is far below the "or lower" score (because they chose far too hard items), so including them would corrupt the statistics.
|Long Test For Genius||3||0.98|
|Qoymans Multiple-Choice #3||5||0.98|
|Analogies of LTFG||3||0.97|
|Spatial Insight Test||3||0.97|
|Cartoons of Shock||4||0.78|
|Space, Time & Hyperspace||4||0.77|
|Genius Association Test||7||0.73|
|Test To End All Tests||5||0.72|
|Qoymans Multiple-Choice #4||7||0.56|
|Test of Shock and Awe||4||0.46|
|Spatial section of Test For Genius Revision 2004||5||0.43|
|Test For Genius - Revision 2004||5||0.3|
|Verbal section of Test For Genius Revision 2004||5||-0.02|
|Association Subtest of LTFG||4||-0.29|
|Cattell Culture Fair||3||-0.65|
Weighted average of correlations: 0.52
Estimated minimum g loading: 0.72
Ranking in above table is based on the unrounded correlations.
|Country||# scores||Average score|
|Father's eduational level||12||0.24|
|Mother's educational level||13||-0.15|
|Disorders (parents and siblings)||13||-0.19|
|Year of birth||13||-0.45|
Correlation of this test with national average IQs published by Lynn and Vanhanen:
To see how the KIT's norming method works, the reported GT scores are equated to IQs using scores on other tests. The GT scale has a high-range mean and SD of 50 and 10, and the KIT's GTs are computed using a method discovered by me, described in the article The Golden Standard of Intelligence. This method does NOT use scores on other tests. So it is interesting to see how its results compare to hypothetical norms based on scores on other tests.
The below IQs are not meant as norms, but to check the working of the new method. Used are scores from other tests correlating .43 and higher with the KIT. The normal method is used (equating ranks of KIT GTs and IQs on other tests)[and the IQs were converted to protonorms using the initial protonorm formula in March 2008].
Although based on only 13 scores, it is worrying that the IQs especially under KIT 55 are much lower than the IQs that ACTUALLY correspond to those GTs when the GTs are computed as explained at the table with IQs, centiles etcetera. For the moment I conclude that this (the KIT's) norming method does not suffice when applied in this manner. To avoid confusion: the height of the norms is unrelated to the standard error of measurement meant on top of this report.