# Performance increase from fish oil?

339 views

### Gwern Branwen

Mar 8, 2010, 9:36:20 PM3/8/10
to N-back
So there was a sale on fish oil at my supermarket, and I grabbed some.
(Turned out to still be a worse price than I could get at Amazon, but
oh well!) I had heard some good things about it in general, and
especially from Seth Godin's blog.

I began taking a fairly high dose: 4 pills daily or 4000mg. I also
used the BW feature to annotate my stats.txt file for when exactly I
started. It's been a week or two, so I took a look at the stats.

Here's the series before I began fish oil. During this time I had the
occasional huperzine-a, but nothing else special (it's roughly 5
scores per day):

[54,69,42,54,44,62,44,35,85,50,44,42,57,65,51,56,42,53,40,47,45,51,57,57,56,76,66,60,46,52,59,48,28,45,43,47,50,40,57,46,33,19,43,58,36,52,44,64]

From the day I began taking the oil:

[55,76,56,55,44,41,44,45,65,70,46,65,46,52,68,52,57,50,64,43,41,50,69,44,47,63,34,57]

At first glance, there's nothing particularly impressive. But if we
look at averages...

> let x = [54,69,42,54,44,62,44,35,85,50,44,42,57,65,51,56,42,53,40,47,45,51,57,57,56,76,66,60,46,52,59,48,28,45,43,47,50,40,57,46,33,19,43,58,36,52,44,64] in sum x `div` length x

53

> let x = [55,76,56,55,44,41,44,45,65,70,46,65,46,52,68,52,57,50,64,43,41,50,69,44,47,63,34,57] in sum x `div` length x

50

3 points is nothing to sneeze at.

Besides the higher average, there's a higher 'floor':

> let x = sort [54,69,42,54,44,62,44,35,85,50,44,42,57,65,51,56,42,53,40,47,45,51,57,57,56,76,66,60,46,52,59,48,28,45,43,47,50,40,57,46,33,19,43,58,36,52,44,64] in take (length x `div` 5) x

[19,28,33,35,36,40,40,42,42]

> let x = sort [55,76,56,55,44,41,44,45,65,70,46,65,46,52,68,52,57,50,64,43,41,50,69,44,47,63,34,57] in take (length x `div` 5) x

[34,41,41,43,44]

One might think the higher floor resulting from less variance (the
series are unequal lengths), but if we add in 'reverse' to look at the
top quintile rather than bottom, we see

[85,76,69,66,65,64,62,60,59]

and

[76,70,69,68,65]

The average of the first is 67, and the second - 69.

So. Not huge differences - nothing that will particularly impress
people - but still a little interesting that some fish oil could boost
my scores a little and make me mentally a little more resistance to
tiring.

I'm not sure it's cost-effective compared to piracetam, but it's
certainly easier to get & take!

--
gwern

### Jonathan Toomim

Mar 8, 2010, 10:47:55 PM3/8/10
Alternate hypothesis: With more practice, you got better.

> --
> You received this message because you are subscribed to the Google
> Groups "Dual N-Back, Brain Training & Intelligence" group.
> To post to this group, send email to brain-t...@googlegroups.com.
> To unsubscribe from this group, send email to brain-trainin...@googlegroups.com
> .
> For more options, visit this group at http://groups.google.com/group/brain-training?hl=en
> .
>

### moeman

Mar 8, 2010, 11:10:41 PM3/8/10
to Dual N-Back, Brain Training & Intelligence
I've started taking fish oil about a week and a half ago and noticed a
slight improvement the first day. A few hours after ingesting 3
1200mg capsules (360mg DHA, 720mg EPA total) I played about 50
sessions of triple-n-back in brain workshop. During the session I got
a perfect score at Triple 6 back (first time occurrence) and made it
to T8B for the first time (been playing triple about 2 weeks prior to
that). It seems as if the increase in performance only lasted for
that particular day but I could be wrong. There has been a
considerable improvement in mood since taking the fish oil so I'm
definitely going to continue taking it.

### Gwern Branwen

Mar 8, 2010, 11:20:33 PM3/8/10
to brain-training
On Mon, Mar 8, 2010 at 10:47 PM, Jonathan Toomim <jto...@jtoomim.org> wrote:
> Alternate hypothesis:  With more practice, you got better.

Obviously, I considered that. I don't think this is the case, as I
have been stuck at D4B since at least September with not too much
change in my score.

I'll take my scores from right before the fish oil to back into
September. The Haskell for sliding averages with window of 10 is:

> map (\x -> sum x `div` length x) \$ takeWhile (\x -> length x >= 10) \$ map (take 10) \$ tails [62,30,60,48,46,52,60,60,48,47,34,50,51,45,54,70,48,61,43,53,60,44,57,50,50,52,37,55,40,53,48,50,52,44,50,50,38,43,66,40,24,67,60,71,54,51,60,41,58,20,28,42,53,59,42,31,60,42,58,36,48,53,46,25,53,57,60,35,46,32,26,68,45,20,51,56,48,25,62,50,54,47,42,55,39,60,44,32,50,34,60,47,70,68,38,47,48,70,51,42,41,35,36,39,23,50,46,44,56,50,39,38,30,51,50,55,60,61,65,43,50,58,52,40,46,58,61,38,50,45,41,45,39,60,38,41,29,29,47,36,39,60,61,38,46,36,36,51,64,52,52,45,42,50,46,66,46,50,63,40,50,50,60,57,57,63,50,31,32,56,20,55,46,35,41,68,44,54,69,42,54,44,62,44,35,85,50,44,42,57,65,51,56,42,53,40,47,45,51,57,57,56,76,66,60,46,52,59,48,28,45,43,47,50,40,57,46,33,19,43,58,36,52,44,64]

The sliding averages:

[51,48,50,49,49,50,51,50,50,50,50,53,52,53,54,53,51,50,50,49,49,48,49,48,48,48,47,48,46,49,48,45,47,48,50,51,51,53,53,52,50,51,48,47,46,45,43,43,43,43,45,47,48,47,44,45,47,47,47,45,45,43,44,44,44,44,43,42,41,43,45,47,45,45,49,47,48,47,48,47,45,46,46,49,50,50,49,49,53,53,54,52,51,47,44,43,43,43,40,41,42,41,42,41,42,45,45,47,49,49,49,50,52,54,53,53,53,53,51,49,49,48,47,46,48,47,45,42,41,41,40,40,41,44,41,42,42,42,45,46,48,49,48,46,47,47,50,51,51,51,50,49,50,52,52,53,53,54,52,49,50,47,48,46,44,42,43,42,45,48,47,50,49,51,52,51,53,53,52,50,51,52,53,52,52,54,50,49,49,50,50,49,50,52,54,55,56,56,58,57,54,53,52,49,47,45,46,46,43,40,42,43,42,43,42,45]

If we broaden the window to 29, the total number of oil-influenced
scores which were being averaged and thus a more fair comparison than
just a window of 10, the averages are:

[50,50,50,50,50,50,50,50,49,49,49,50,49,49,50,50,50,50,50,50,50,49,48,48,48,48,48,47,48,48,48,47,47,47,47,47,47,47,48,47,47,47,46,46,45,44,44,44,44,43,45,45,46,46,45,45,46,46,46,45,45,45,45,45,47,47,47,46,47,47,48,49,48,47,48,47,46,46,47,46,47,47,46,46,45,46,45,46,47,47,48,48,48,47,47,47,47,47,47,46,47,47,47,47,47,49,48,48,47,46,46,46,46,47,48,47,47,46,45,45,46,46,45,45,45,45,45,45,45,45,46,46,46,47,47,47,48,49,50,49,49,50,48,48,48,48,48,49,49,49,49,49,49,49,50,49,48,50,50,49,49,49,50,50,50,49,49,48,49,49,49,50,51,51,52,53,53,53,53,53,53,52,52,51,51,52,50,51,51,50,49,48,48,48,48,48,49]

Note how few there are that even equal the 53 of the oil average;
there is no pre-oil 29-average which is >53.

If you are skeptical, feel free to look at the stats yourself.

--
gwern

stats.txt

### Enoch Alien

Mar 9, 2010, 12:20:56 PM3/9/10
Moeman,

Are you the same dude that's got up to level five on cognitivefun
memory updating game? Test 3?

Message has been deleted

### Jonathan Toomim

Mar 9, 2010, 1:28:30 PM3/9/10
I performed a two-sample t-test.  The data before and after fish oil supplementation do not come from distributions with significantly different means (p > .23).  In other words, the data you presented are better explained by chance than by fish oil, at least according to a two-sample t-test.

### Gwern Branwen

Mar 9, 2010, 1:44:42 PM3/9/10
to brain-training
On Tue, Mar 9, 2010 at 1:28 PM, Jonathan Toomim <jto...@jtoomim.org> wrote:
> I performed a two-sample t-test.  The data before and after fish oil
> supplementation do not come from distributions with significantly different
> means (p > .23).  In other words, the data you presented are better
> explained by chance than by fish oil, at least according to a two-sample
> t-test.

Could you explain that test? I'm not much of a statistician.

--
gwern

Message has been deleted

### wzeller

Mar 9, 2010, 2:04:58 PM3/9/10
On Mar 9, 2010, at 12:28 PM, Jonathan Toomim wrote:

I performed a two-sample t-test.  The data before and after fish oil supplementation do not come from distributions with significantly different means (p > .23).  In other words, the data you presented are better explained by chance than by fish oil, at least according to a two-sample t-test.

I disagree with your final sentence.  All you can say from that p-value is that the null hypothesis, or the hypothesis that any difference between the data from before and from after fish oil supplementation is due to chance alone, cannot be ruled out with the degree of confidence ordinarily required by science to demonstrate an effect.  Specifically, that p-value says that the chance of obtaining both the pre- and post-fish oil data sets from the same data set is .23.

It does not follow, however, that chance is a better explanation for the difference between the two data sets than the presence or absence of fish oil.  Absence of evidence is not evidence of absence (see http://en.wikipedia.org/wiki/Argument_from_ignorance).  From the p-value you calculated, there is a 77% chance that the pre- and post-fish oil scores were meaningfully different.  If anything, any p-value under .5 means that while the more likely explanation is that the two data sets are different, the difference is not extreme enough to say that they are, in fact, different with the required confidence (p < .05 or .01).

On Mar 8, 2010, at 8:20 PM, Gwern Branwen wrote:

On Mon, Mar 8, 2010 at 10:47 PM, Jonathan Toomim <jto...@jtoomim.org> wrote:
Alternate hypothesis:  With more practice, you got better.

Obviously, I considered that. I don't think this is the case, as I
have been stuck at D4B since at least September with not too much
change in my score.

### Paul Hobbs

Mar 9, 2010, 2:16:33 PM3/9/10
Also, remember the expected result is that it would improve the score,
so using Baye's law, this increases the support for our hypothesis
that fish oil helps.
--
Paul Hobbs

### Jonathan Toomim

Mar 9, 2010, 2:43:13 PM3/9/10
Also, remember that (a) this was not a controlled experiment, and that
people tend to improve on things they practice at, (b) it was not a
blind experiment, (c) reporting bias is almost certainly present, (d)
omega-3s (at least, DHA and EPA) are believed to require weeks or
months before any effect is seen, whereas placebo effects peak early
and fade with time, and (e) the "improvement" that was seen peaked
shortly after beginning omega-3 consumption and declined quickly
thereafter.

There are plenty of factors at play here which could explain the
purported effect seen other than omega-3s, and the temporal
characteristics of the effect are not consistent with current
understanding of how omega-3s work, so it is not parsimonious to
ascribe the effect to omega-3s. Pluritas non est poneda sine
necessitate.

### moeman

Mar 9, 2010, 2:53:01 PM3/9/10
to Dual N-Back, Brain Training & Intelligence
On Mar 9, 12:20 pm, Enoch Alien <alienenoch...@googlemail.com> wrote:
> Moeman,
>
> Are you the same dude that's got up to level five on cognitivefun
> memory updating game? Test 3?

Actually I made it up to level 6 yesterday but struck out hard once I
got there.

On Mar 9, 12:20 pm, Enoch Alien <alienenoch...@googlemail.com> wrote:
> Moeman,
>
> Are you the same dude that's got up to level five on cognitivefun
> memory updating game? Test 3?
>

### wzeller

Mar 9, 2010, 3:00:33 PM3/9/10
I'm not trying to get into an argument here. I really value your input to the group and agree with (and find extremely interesting) nearly all of your posts. I am only saying than your statement, which was that "the data you presented are better explained by chance than by fish oil, at least according to a two-sample t-test," is false.

On Mar 9, 2010, at 1:43 PM, Jonathan Toomim wrote:

> Also, remember that (a) this was not a controlled experiment, and that people tend to improve on things they practice at, (b) it was not a blind experiment, (c) reporting bias is almost certainly present, (d) omega-3s (at least, DHA and EPA) are believed to require weeks or months before any effect is seen, whereas placebo effects peak early and fade with time, and (e) the "improvement" that was seen peaked shortly after beginning omega-3 consumption and declined quickly thereafter.
>
> There are plenty of factors at play here which could explain the purported effect seen other than omega-3s, and the temporal characteristics of the effect are not consistent with current understanding of how omega-3s work, so it is not parsimonious to ascribe the effect to omega-3s. Pluritas non est poneda sine necessitate.
>
> On Mar 9, 2010, at 11:16 AM, Paul Hobbs wrote:
>
>> Also, remember the expected result is that it would improve the score,
>> so using Baye's law, this increases the support for our hypothesis
>> that fish oil helps.
>> --
>> Paul Hobbs
>> On Tue, Mar 9, 2010 at 11:04 AM, wzeller <wze...@gmail.com> wrote:
>>>
>>> On Mar 9, 2010, at 12:28 PM, Jonathan Toomim wrote:
>>>
>>> I performed a two-sample t-test. The data before and after fish oil
>>> supplementation do not come from distributions with significantly different
>>> means (p > .23). In other words, the data you presented are better
>>> explained by chance than by fish oil, at least according to a two-sample
>>> t-test.
>>>

>>> I disagree with your final sentence. All you can say from that p-value is

>> To unsubscribe from this group, send email to brain-trainin...@googlegroups.com.

Message has been deleted
Message has been deleted

### moeman

Mar 9, 2010, 6:50:37 PM3/9/10
to Dual N-Back, Brain Training & Intelligence
> Moeman,
> So what's the trick? I'm actually up to level 5 but can't do it
> simultaneously.

On level 5 simultaneous I initially try to keep track of all 5 boxes
but one usually drops out of my mind so I end up keeping track of 4
most of the time. I love playing simultaneously as it's a heck of a
workout.

On Mar 9, 5:07 pm, Enoch Alien <alienenoch...@googlemail.com> wrote:
> Moeman,

> So what's the trick? I'm actually up to level 5 but can't do it
> simultaneously.

### jaBuddha

Mar 10, 2010, 3:50:39 PM3/10/10
to Dual N-Back, Brain Training & Intelligence

"Clearly, omega-3 fatty acids are essential to good brain health."
In: Healing ADD: "The Breakthrough Program That Allows You to See and
Heal
the 6 Types of ADD. P. 263." By: Daniel G. Amen, M.D. See: www.brainplace.com

Jared R. Goldman
School Psychologist
EE/SS Northeast Elementary
Ph: 904-858-6176

On Mar 8, 11:10 pm, moeman <lindasmit...@bellsouth.net> wrote:
> I've started taking fish oil about a week and a half ago and noticed a
> slight improvement the first day.  A few hours after ingesting 3
> 1200mg capsules (360mg DHA, 720mg EPA total) I played about 50
> sessions of triple-n-back in brain workshop.  During the session I got
> a perfect score at Triple 6 back (first time occurrence) and made it
> to T8B for the first time (been playing triple about 2 weeks prior to
> that).  It seems as if the increase in performance only lasted for
> that particular day but I could be wrong.  There has been a
> considerable improvement in mood since taking the fish oil so I'm
> definitely going to continue taking it.
>
> On Mar 8, 9:36 pm, Gwern Branwen <gwe...@gmail.com> wrote:
>
>
>
> > So there was a sale on fish oil at my supermarket, and I grabbed some.
> > (Turned out to still be a worse price than I could get at Amazon, but
> > oh well!) I had heard some good things about it in general, and
> > especially from Seth Godin's blog.
>
> > I began taking a fairly high dose: 4 pills daily or 4000mg. I also
> > used the BW feature to annotate my stats.txt file for when exactly I
> > started. It's been a week or two, so I took a look at the stats.
>
> > Here's the series before I began fish oil. During this time I had the
> > occasional huperzine-a, but nothing else special (it's roughly 5
> > scores per day):
>

> > [54,69,42,54,44,62,44,35,85,50,44,42,57,65,51,56,42,53,40,47,45,51,57,57,56­,76,66,60,46,52,59,48,28,45,43,47,50,40,57,46,33,19,43,58,36,52,44,64]

>
> > From the day I began taking the oil:
>

> > [55,76,56,55,44,41,44,45,65,70,46,65,46,52,68,52,57,50,64,43,41,50,69,44,47­,63,34,57]

>
> > At first glance, there's nothing particularly impressive. But if we
> > look at averages...
>

> > > let x = [54,69,42,54,44,62,44,35,85,50,44,42,57,65,51,56,42,53,40,47,45,51,57,57,56­,76,66,60,46,52,59,48,28,45,43,47,50,40,57,46,33,19,43,58,36,52,44,64]  in sum x `div` length x
>
> > 53
>
> > > let x = [55,76,56,55,44,41,44,45,65,70,46,65,46,52,68,52,57,50,64,43,41,50,69,44,47­,63,34,57]  in sum x `div` length x

>
> > 50
>
> > 3 points is nothing to sneeze at.
>
> > Besides the higher average, there's a higher 'floor':
>

> > > let x = sort [54,69,42,54,44,62,44,35,85,50,44,42,57,65,51,56,42,53,40,47,45,51,57,57,56­,76,66,60,46,52,59,48,28,45,43,47,50,40,57,46,33,19,43,58,36,52,44,64] in take (length x `div` 5) x
>
> > [19,28,33,35,36,40,40,42,42]
>
> > > let x = sort [55,76,56,55,44,41,44,45,65,70,46,65,46,52,68,52,57,50,64,43,41,50,69,44,47­,63,34,57] in take (length x `div` 5) x

>
> > [34,41,41,43,44]
>
> > One might think the higher floor resulting from less variance (the
> > series are unequal lengths), but if we add in 'reverse' to look at the
> > top quintile rather than bottom, we see
>
> > [85,76,69,66,65,64,62,60,59]
>
> > and
>
> > [76,70,69,68,65]
>
> > The average of the first is 67, and the second - 69.
>
> > So. Not huge differences - nothing that will particularly impress
> > people - but still a little interesting that some fish oil could boost
> > my scores a little and make me mentally a little more resistance to
> > tiring.
>
> > I'm not sure it's cost-effective compared to piracetam, but it's
> > certainly easier to get & take!
>
> > --

> > gwern- Hide quoted text -
>
> - Show quoted text -