Took the survey, plus the IQ test out of curiosity, I’d never had my IQ tested before.
Along similar reasoning, do we know how well the iqtest.dk test correlates with non-internet tests of IQ? Getting a number is cool, knowing it was generated by a process fundamentally different than rand(100,160) would be even better
I strongly suspect that a lot of the members of LessWrong have had a non-internet IQ test and will have entered their scores on the census. Those who also took the extra credit internet test and entered their scores to that as well could serve as a sample group for us to make just such an analysis.
Granted, we are likely a biased sample of the population (I suspect a median of somewhere around 125 for both tests), but data is data.
From what I could read on the iqtest page, it seemed that they didn’t do any correction for self-selection bias, but rather calculated scores as if they had a representative sample. Based on this I would guess that the internet IQ test will underestimate your score (p=0.7)
Unless there are significant numbers of people, myself for example, who take the test multiple times with varied random algorithms just to see how it affects the outcome. I’d only put a (p=0.55) at the test underestimating your score, conditional that it doesn’t correct for self-selection bias.
Though, given that the lowest score appears to be “less than 79”, rather than an exact number, they may simply drop any scores under 79 from their pool, or at the very least weight them differently. Has anybody identified a similar maximum score which would support this hypothesis of discarding outliers?
A lot of people in this thread have reported institutionally tested IQ scores much higher than those given by the extra-credit Internet test. (My own score on the former is two and a half standard deviations higher than on the latter, though I took the former many years ago.) I suspect that iqtest.dk’s either normed very low (unusually for an Internet IQ test) or is suffering from other problems.
The first possibility that comes to mind is that some people are taking it several times (one person elsewhere in the thread reported scores in the 100 range, then 120, then 140 over three trials), and that its scoring system is taking that into account improperly.
As I mention elsewhere, … gur grfg vf vaperqvoyl rnfl gb tnzr. Vg qbrfa’g punatr gur cbfvgvba bs cbgragvny nafjref orgjrra gnxvatf, fb trggvat n cresrpg fpber ba lbhe frpbaq gel vf cerggl rnfl.
That’s what I thought. So if jeremysalwen is right and they’re calculating scores as if their sample of results is representative, then their scoring algorithm is going to be utterly useless (perhaps outside a narrow regime around 100): the right-hand side of the curve is going to be populated entirely by people gaming the test in various ways (retakes are almost certainly common, and more sophisticated methods are likely to represent at least a couple percent of results), and I’d imagine a good chunk of the left-hand side would fall to guessers.
I assumed that iqtest.dk would count as “respectable” rather than as “amateur” so I included its score in both questions. If you do such a calibration, you should ignore my entry.
Took the survey, plus the IQ test out of curiosity, I’d never had my IQ tested before.
Along similar reasoning, do we know how well the iqtest.dk test correlates with non-internet tests of IQ? Getting a number is cool, knowing it was generated by a process fundamentally different than rand(100,160) would be even better
I strongly suspect that a lot of the members of LessWrong have had a non-internet IQ test and will have entered their scores on the census. Those who also took the extra credit internet test and entered their scores to that as well could serve as a sample group for us to make just such an analysis.
Granted, we are likely a biased sample of the population (I suspect a median of somewhere around 125 for both tests), but data is data.
From what I could read on the iqtest page, it seemed that they didn’t do any correction for self-selection bias, but rather calculated scores as if they had a representative sample. Based on this I would guess that the internet IQ test will underestimate your score (p=0.7)
Unless there are significant numbers of people, myself for example, who take the test multiple times with varied random algorithms just to see how it affects the outcome. I’d only put a (p=0.55) at the test underestimating your score, conditional that it doesn’t correct for self-selection bias.
Though, given that the lowest score appears to be “less than 79”, rather than an exact number, they may simply drop any scores under 79 from their pool, or at the very least weight them differently. Has anybody identified a similar maximum score which would support this hypothesis of discarding outliers?
Analysis of the survey results seems to indicate that I was correct: http://lesswrong.com/lw/fp5/2012_survey_results/
A lot of people in this thread have reported institutionally tested IQ scores much higher than those given by the extra-credit Internet test. (My own score on the former is two and a half standard deviations higher than on the latter, though I took the former many years ago.) I suspect that iqtest.dk’s either normed very low (unusually for an Internet IQ test) or is suffering from other problems.
The first possibility that comes to mind is that some people are taking it several times (one person elsewhere in the thread reported scores in the 100 range, then 120, then 140 over three trials), and that its scoring system is taking that into account improperly.
As I mention elsewhere, … gur grfg vf vaperqvoyl rnfl gb tnzr. Vg qbrfa’g punatr gur cbfvgvba bs cbgragvny nafjref orgjrra gnxvatf, fb trggvat n cresrpg fpber ba lbhe frpbaq gel vf cerggl rnfl.
That’s what I thought. So if jeremysalwen is right and they’re calculating scores as if their sample of results is representative, then their scoring algorithm is going to be utterly useless (perhaps outside a narrow regime around 100): the right-hand side of the curve is going to be populated entirely by people gaming the test in various ways (retakes are almost certainly common, and more sophisticated methods are likely to represent at least a couple percent of results), and I’d imagine a good chunk of the left-hand side would fall to guessers.
Even without cheating, I don’t think the people who voluntarily spend 40 minutes on an IQ test are representative of the whole population.
I assumed that iqtest.dk would count as “respectable” rather than as “amateur” so I included its score in both questions. If you do such a calibration, you should ignore my entry.