a very unofficial dnb-iq research
polar
well, thanks to Paul, volunteers and my uni faculty, I eventually
finished some... data. Everything was done without any dotation, so
half of the volunteers just "ceased to be" :) Now, in group of 20
people (which trained on their own) there's no point to be more
precise than just years of age, minutes of test time, or to make any
sophisticated statistics. But any feedback is welcomed, or if you
want, just copy everything to new .xls and do your stats. So here's
the link, enjoy if possible..
http://spreadsheets.google.com/ccc?key=rOfijmsJ-hxPbzmbi4dmHVg
Gwern Branwen
Hash: SHA512
On Wed, Jun 17, 2009 at 1:02 PM, polar wrote:
> Hello,
>
> well, thanks to Paul, volunteers and my uni faculty, I eventually
> finished some... data. Everything was done without any dotation, so
> half of the volunteers just "ceased to be" :)
That's quite some shrinkage!
> Now, in group of 20
> people (which trained on their own) there's no point to be more
> precise than just years of age, minutes of test time, or to make any
> sophisticated statistics. But any feedback is welcomed, or if you
> want, just copy everything to new .xls and do your stats. So here's
> the link, enjoy if possible..
>
> http://spreadsheets.google.com/ccc?key=rOfijmsJ-hxPbzmbi4dmHVg
Hm. It's a little odd that your control group is so disparate and
small compared to the test group. But let's eyeball the results.
Omitting the 2 worthless ones, we see the control group improves by
3,1,2,0,0,4, for an average of ~1.6, or let's call it 2.
The test group improves by 2,1,2,4,3,0,4,2,2,6,4,4,2, for an average
of ~2.7, or 3.
And you note that 8/12 of the testing group did faster & better,
compared to 1 of the control group (I don't understand your annotation
about a 'mistake'.)
The training seems a little odd; they did just six hours over 5 weeks,
or 10 minutes a day? Other people here recommend at least double that.
I wonder whether the difference would've been more substantial.
Small study, and small effect, but I suppose it fits in with what we expect.
- --
gwern
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
iEYEAREKAAYFAko5JzgACgkQvpDo5Pfl1oKLcwCfUOHq89CJOsb5guXjOCuFuLXN
f7kAnA+tj/CzpWeTdHit1jmb0O+QWkwd
=GJcZ
-----END PGP SIGNATURE-----
Martin Syk
statistical measures) it seems as if the dual-n-backers generally
improved a couple of points on RPM and became quicker (which may be a
training effect). The big weaknesses are the size of the control group
and the quality of their results... Without controls training with
another type of material (for instance crosswords) it is hard to know
what explains the effects.
Nevertheless, thank you for doing this. Did you collect any
information about how the training group estimated their abilities
before/after or if the noticed any change?
Pontus Granström
Mike L.
this would be equivalent to only 12 days of training, according to the
Jaeggi study in which the participants trained for 30 minutes each day
for 19 days. which would in fact be equivocal to 9.5 hours of
training.
Mike L.
note that most of the gain in so-called intelligence came between
12-19 days (gain in int. went from ~1.5 to ~4.5).
This might actually be important to look at for future studies;
perhaps the gain in intelligence comes more and more prominently after
having trained with DNB for longer periods of time.
In any case, the above mentioned is important to note because the
study conducted by polar trained its participants for the equivalent
of 12 days, not 19 (as i mentioned in an earlier post).
Vlad
yeah, it's a pitty that I could not really choose the people, their
number or help them with motivation (altough i tried). Of course, my
instructions were to train for 19 days, always 20 sessions, but they
omitted sometimes some days, and several people afterwards admitted
that they did less than 19 regular days. And, the second testing did
not take place immediately, but on average 1 week after the training.
But I got few, very clear confirmations about subjective feelings of
alertness, better memory and thinking during and after the training.
From my personal experience I agree with them, and from professional
point of view (am a psychologist) find it slightly encouraging.
Mike L.
Mike L.
that the difference between the control/test group's average improved
test scores might increase in size (that is, grow further apart) with
time....
so, for example, if right now the average difference between the two
scores (1.6, 2.7) is 1.1, and that was achieved after 6 hours of
training...i think it's safe to hypothesize that if the time spent
training increased, so then, would the effective increase in score
(for the group that trains) as well as the difference between the
control group and the test group's scores.
we might see something like:
[after 24 hours of training] Average improvement in control group test
scores: 1.8 points //\\ Average improvement in test group scores: 5.8
(a 4 point difference now, as opposed to a 1.1 point difference for
the test group that trained 6 hours)
Vlad
I see your point about the iq gain in later days of training (and
actually agree), and in possible widening the differences with time.
Reason #1 why I am cautious about interpreting results, is not the
outdated RAPM II test, or even small groups, but the fact, that each
point of raw score has different "weight". Participants usually do
like 2/3 of questions in 1/3 time (...), but every question equals 1
point in the end. So it's a big difference if you do 4 mistakes in the
middle of the test (and get the 6 hardest right), or you have first 31
right, but then spent 15 minutes without succes on question 32...
Usually, people wrote all answers, and they had max. 3 errors from
last 7 questions - but at the same time, they always had at least 3
errors before question 30.
Reason number #2 is we cant know how much time exactly did they train.
Paul even made a new feature (encrypted logfile), but most people did
not send it to me at the end for control. It's very easy to be
motivated for "increasing intelligence" first three days, but after
two weeks people get very frustrated. Altough this is a pity, I think
this is legitimate topic for research, because longer-lasting changes
in brain usually takes 2-3 weeks of something (medication for
example).
I think DNB works pretty good, and (at least temporary) can
significantly increase fluid intelligence. But this study was only
first (and kind of trivial) attempt. There's need for far better
research (which I'll try as soon as possible :). An I support every
attempt for research made here.