The Effect of Training Working Memory and Attention on Pupils’ Fluid Intelligence
472 views
Skip to first unread message
XFMQ902SF
Jun 29, 2012, 4:10:40 PM6/29/12
to Dual N-Back, Brain Training & Intelligence
This has to be the most poorly written abstract ever, however, the
study described includes n-back and claims to provide evidence of
transfer of working memory training to g.
Working memory, attention and intelligence are closely related, and
some studies suggest that intelligence can be used as an indicator of
prediction, but whether can improve fluid intelligence, research
results less consistent; Whether the improvement of fluid intelligence
caused by training of working memory related to training task is not
clear; The attention training impact on fluid intelligence is not yet
clear; In addition, from children training point of view. there is not
much research literature. This paper mainly discusses the training
effect on fluid intelligence and working memory from the angle of
children working memory training and attention training.This study
includes three experiments. The main purpose of preliminary
experiments is to understand the subjects performance on the the
single task of visual working memory, working memory dual n-back task
and dual-channel attention task at the same time.Then, according to
the performance of the subjects, we will make a standard which decides
the next block difficulty of subjects. We made the difficulty standard
according to the performance of the subjects:when the number of
correct trial of each block t between 13-18, n constant in the next
block; more than or equal to 19, n+1 back in the next block; less than
equal to 12, n-1back in the next block.The performance on Single
visual working memory task and dual n-back working memory task is in
line with expectations, we can distinguish between the four levels of
difficulty; There is not a increases difficulty on dual-channel
attention task between the four levels, so the task requires further
improvement.The purpose of the first experiment is to explore whether
WM training and attention training can improve GF; the influence of
different training mission; the effect of mount of training on the
improvement of GF; whether the rise of GF would improve children’s
academic achievement.We expect:training can improve fluid intelligence
and working memory span; different training tasks have a different
effect on improving fluid intelligence, only dual n-back working
memory training can improve fluid intelligence; the performance of
working memory training and attention training tasks will increase
with training amount; Ideally, the training effect will be reflected
in the children’s academic achievement and improvement.We have trained
and control subjects who all accepted pre-test and post-test of fluid
intelligence,working memory reading span and working memory digit span
before and after the training.The trained subjects accepted either
working memory training or attention training,only one task each, and
meanwhile, the control group didn’t do anything. It should be pointed
out that there is difference in the amount of training between the
trained groups. There are four training levels (five days, ten days,
fifteen days, twenty days) of training amount.There are five major
research findings:(1) The training gain was well explained by a linear
function for every group of working memory and attention training
tasks. (2) Fifteen days of training is best to improve fluid
intelligence and working memory reading span. (3) The simultaneous
double n-back working memory task is the best task to increase fluid
intelligence. (4) The ANOVA and Covariance ANOVA found that
participants with initially lower fluid intelligence, working memory
reading and digit span showed even larger gains on them.(5)Training is
good for improving math scores.The goal of experiment 2 is to explore
whether 15 days of training has a special role on improving fluid
intelligence; The benefit of low group on fluid intelligence, reading
span, digital span is greater than the high group. Experiment 2
hypothesis:Fifteen days of training have special effect to improve
fluid intelligence; The low group get benefit more than the high
group.The first independent variable includes15 days and20 days two
level; The second independent variable includes training, control two
level. The other conditions are the same as experiment one. Experiment
two shows that fluid intelligence group significantly higher than
control group; There isn’t significant differences between 15 days and
19 days on the influence of fluid intelligence; The benefit of low
group was greater than the high group on fluid intelligence、reading
span and digital span.Integrating experiment one and two, this
research shows that:(1) The training of working memory or attention
can improve fluid intelligence, simultaneous double n-back WM task was
the best to improve fluid intelligence; The Training group was
significantly better than control group on fluid intelligence、WM span
and math scores.(2) The training gain was well explained by a linear
function for every group of working memory and attention training
tasks. But there is no significant role of training amount. (3)
Whether on fluid intelligence, working memory span or Chinese、math
scores, the benefit of low group was obviously higher than the high
group.
http://www.p-papers.com/22515.html
study described includes n-back and claims to provide evidence of
transfer of working memory training to g.
Working memory, attention and intelligence are closely related, and
some studies suggest that intelligence can be used as an indicator of
prediction, but whether can improve fluid intelligence, research
results less consistent; Whether the improvement of fluid intelligence
caused by training of working memory related to training task is not
clear; The attention training impact on fluid intelligence is not yet
clear; In addition, from children training point of view. there is not
much research literature. This paper mainly discusses the training
effect on fluid intelligence and working memory from the angle of
children working memory training and attention training.This study
includes three experiments. The main purpose of preliminary
experiments is to understand the subjects performance on the the
single task of visual working memory, working memory dual n-back task
and dual-channel attention task at the same time.Then, according to
the performance of the subjects, we will make a standard which decides
the next block difficulty of subjects. We made the difficulty standard
according to the performance of the subjects:when the number of
correct trial of each block t between 13-18, n constant in the next
block; more than or equal to 19, n+1 back in the next block; less than
equal to 12, n-1back in the next block.The performance on Single
visual working memory task and dual n-back working memory task is in
line with expectations, we can distinguish between the four levels of
difficulty; There is not a increases difficulty on dual-channel
attention task between the four levels, so the task requires further
improvement.The purpose of the first experiment is to explore whether
WM training and attention training can improve GF; the influence of
different training mission; the effect of mount of training on the
improvement of GF; whether the rise of GF would improve children’s
academic achievement.We expect:training can improve fluid intelligence
and working memory span; different training tasks have a different
effect on improving fluid intelligence, only dual n-back working
memory training can improve fluid intelligence; the performance of
working memory training and attention training tasks will increase
with training amount; Ideally, the training effect will be reflected
in the children’s academic achievement and improvement.We have trained
and control subjects who all accepted pre-test and post-test of fluid
intelligence,working memory reading span and working memory digit span
before and after the training.The trained subjects accepted either
working memory training or attention training,only one task each, and
meanwhile, the control group didn’t do anything. It should be pointed
out that there is difference in the amount of training between the
trained groups. There are four training levels (five days, ten days,
fifteen days, twenty days) of training amount.There are five major
research findings:(1) The training gain was well explained by a linear
function for every group of working memory and attention training
tasks. (2) Fifteen days of training is best to improve fluid
intelligence and working memory reading span. (3) The simultaneous
double n-back working memory task is the best task to increase fluid
intelligence. (4) The ANOVA and Covariance ANOVA found that
participants with initially lower fluid intelligence, working memory
reading and digit span showed even larger gains on them.(5)Training is
good for improving math scores.The goal of experiment 2 is to explore
whether 15 days of training has a special role on improving fluid
intelligence; The benefit of low group on fluid intelligence, reading
span, digital span is greater than the high group. Experiment 2
hypothesis:Fifteen days of training have special effect to improve
fluid intelligence; The low group get benefit more than the high
group.The first independent variable includes15 days and20 days two
level; The second independent variable includes training, control two
level. The other conditions are the same as experiment one. Experiment
two shows that fluid intelligence group significantly higher than
control group; There isn’t significant differences between 15 days and
19 days on the influence of fluid intelligence; The benefit of low
group was greater than the high group on fluid intelligence、reading
span and digital span.Integrating experiment one and two, this
research shows that:(1) The training of working memory or attention
can improve fluid intelligence, simultaneous double n-back WM task was
the best to improve fluid intelligence; The Training group was
significantly better than control group on fluid intelligence、WM span
and math scores.(2) The training gain was well explained by a linear
function for every group of working memory and attention training
tasks. But there is no significant role of training amount. (3)
Whether on fluid intelligence, working memory span or Chinese、math
scores, the benefit of low group was obviously higher than the high
group.
http://www.p-papers.com/22515.html
Gwern Branwen
Jun 29, 2012, 4:41:39 PM6/29/12
to brain-t...@googlegroups.com
A better link is http://www.globethesis.com/?t=2155330335456903 which
tells us that the absence of any journal information is because it's a
Master's thesis by one C J Zhong.
Getting it may be hard; I don't see any university identified or even
whether it's in English (sometimes just the abstract is English and
the rest is the native language).
Incidentally, all 4 authors of Qiu et al 2009 have yet to reply to my
request for the actual data despite multiple emails over the last
month or two. I reiterate my concerns about Chinese research.
--
gwern
http://www.gwern.net
Gwern Branwen
Jun 29, 2012, 9:15:25 PM6/29/12
to brain-t...@googlegroups.com
On Fri, Jun 29, 2012 at 4:10 PM, XFMQ902SF <kei...@aol.com> wrote:
> The trained subjects accepted either
> working memory training or attention training,only one task each, and
> meanwhile, the control group didn’t do anything.
This seems like an important point buried in the abstract: this is
> working memory training or attention training,only one task each, and
> meanwhile, the control group didn’t do anything.
another no-contact control group, and there's no mention of the other
experiments using active controls. So it sounds like this study would
only increase the split in effect size between the no-contact and
active studies.
--
gwern
http://www.gwern.net
Gwern Branwen
Jun 29, 2012, 11:00:52 PM6/29/12
to brain-t...@googlegroups.com
Good news, this paper is available online:
http://www.doc88.com/p-397166703921.html
Bad news, it's all Chinese except the abstract. I scrolled through it,
hoping to find a table which looked like RAPM scores for all groups,
but unfortunately, there turn out to be quite a few tables! I can't
tell which is which.
Are there any Chinese readers here who would be willing to pull out
the right numbers for me? For the meta-analysis, I need number of
subjects in each group, and the IQ scores before-after with
SD/standard-deviations.
--
gwern
http://www.gwern.net
http://www.doc88.com/p-397166703921.html
Bad news, it's all Chinese except the abstract. I scrolled through it,
hoping to find a table which looked like RAPM scores for all groups,
but unfortunately, there turn out to be quite a few tables! I can't
tell which is which.
Are there any Chinese readers here who would be willing to pull out
the right numbers for me? For the meta-analysis, I need number of
subjects in each group, and the IQ scores before-after with
SD/standard-deviations.
--
gwern
http://www.gwern.net
XFMQ902SF
Jun 30, 2012, 12:34:49 AM6/30/12
to Dual N-Back, Brain Training & Intelligence
I believe Jausovec 2012 had an active control group with plenty of
transfer later on.
On Jun 29, 9:15 pm, Gwern Branwen <gwe...@gmail.com> wrote:
transfer later on.
On Jun 29, 9:15 pm, Gwern Branwen <gwe...@gmail.com> wrote:
Gwern Branwen
Jun 30, 2012, 11:29:16 AM6/30/12
to brain-t...@googlegroups.com
On Sat, Jun 30, 2012 at 12:34 AM, XFMQ902SF <kei...@aol.com> wrote:
> I believe Jausovec 2012 had an active control group with plenty of
> transfer later on.
With a weird testing procedure, and it's one of the few to show much.
> I believe Jausovec 2012 had an active control group with plenty of
> transfer later on.
But more importantly, are there any Chinese readers here? All the
numbers seem to be provided, I just need to know what is what...
Working with a Chinese-reading acquaintance on IRC, this is what I have learned:
< MrEmile> lucidian: considerign the (relative) success of things like
siri and Google Translate, I expect that there has been a good deal of
progress these last few years
<gwern> MrEmile: the pg24 4.5.4 table could be IQ before/after
scores... but I have no idea. I think pg25 & 26 probably isn't right
because the numbers are too high
<gwern> pg30 is another possibility
<gwern> MrEmile: as is pg31, now that I check
< MrEmile> gwern: are the page numbers you refer to those on the
paper, or those of the web reader?
<gwern> MrEmile: the web reader itself
< lucidian> MrEmile: I'm just having trouble understanding what it
would mean for a problem to be solved. It seems like accuracy could
always get better, unless we manage to make a tool that gets 100%
accuracy on everything, which is never going to happen.
<gwern> MrEmile: I can't read the page numbers in the paper itself :)
< lucidian> MrEmile: I mean, there does seem to be a point beyond
which further effort is a waste of time, but still.
<gwern> lucidian: since humans don't understand perfectly either,
there's an upper bound!
< MrEmile> gwern: the columns are before (M), before (SD) (I assume
mean and standard deviation), after (M), after (SD)
<gwern> MrEmile: yeah, I guessed as much, but I don't know what the
rows indicate
<gwern> MrEmile: the 5/10/etc probably indicate # of training
days/experimental group
<gwern> MrEmile: but why are there 3 rows for each group? I have no idea
<gwern> 3 or more
< MrEmile> gwern: (on page 24 btw) the lines are (5 days): visual
single task, simultaneous double task, double-sense (I assume sight
and sound) attention task, and control task
< MrEmile> (and the same for 10, 15 and 20 days)
<gwern> MrEmile: hm. that's not what I was expecting. I can guess that
visual single task = single n-back, and double-sense = dual n-back,
but control task seems to imply they did something and I don't know
what simultaneous double task or attention task would be
< MrEmile> gwern: better names would be "single task of visual working
memory", "working memory dual n-back", and "dual-channel attention
task"
< MrEmile> (those are the terms used in the abstract, and the Chinese
abstract uses the terms in the table at the same place)
<gwern> MrEmile: after checking the abstract, looks like the
dual-channel thing is some new task I haven't heard of. I guess it's
not important
< MrEmile> Anything else you need?
<gwern> MrEmile: so that just leaves the next two tables with the 90s
- what are those? - and how many participants they have in each group
<gwern> MrEmile: I've been scanning the text hoping for some
convenient n=x, but no dice so far
< MrEmile> table 5 (with the 90s) looks like success rates on an
english test, before and after the training
<gwern> MrEmile: ah, that must be the reading comprehension test
<gwern> but in a foreign language - english - scores? well, I guess
that makes sense in china
<gwern> (numbers make sese. 90/100 etc, no SDs putting anyone over 100)
< MrEmile> nope
< MrEmile> got it werong
< MrEmile> What they did is have students go through the major exam
before and after training
< MrEmile> table 5 is language (not english), table 6 is maths
<gwern> 'major exam'?
< MrEmile> it's like the exam to get into university
< MrEmile> I dunno the english equivalent
<gwern> ah, so the big end of highschool thing
< MrEmile> I assume it was a simulation of the exam, though they don't say
< MrEmile> yeah
<gwern> I was assuming these were like elementary school kids
< MrEmile> They don't specifiy the year, actually it's the exam you
always get at the end of a year to go to the next one
< MrEmile> they don't specify the age either
< MrEmile> (well, they probably do, somewhere else :P)
<gwern> so are pg23/24 basically the same data, reporting IQ scores
before/after, just with different subjects? I'm otherwise confused why
the two tables look so identical
<gwern> I almost hope they aren't because that makes finding the
subject number even more difficult - 4 groups including control, in 2
experiments, is 8 separate groups, and then I'd need to pool their
means and standad deviations
< MrEmile> table 3 is remembering words, table 4 is remembering
numbers apparently
<gwern> ;_; so where are the damn IQ scores
< MrEmile> Dunno. There doesn't seem to be any in any table
<gwern> table 3 is verbal WM, table 4 is digit span, table 5 is
language scores, table 6 is math scores; so that leaves table 2 on
pg22 and table 7 & 8 on pg30
< MrEmile> actually table 2 is fluid intelligence
<gwern> 8's numbers look way too small
<gwern> MrEmile: oh, great.
<gwern> you already translated the 4 groups, so I guess the last
remaining detail is how many kids were in each of the 4 - the single &
dual n-back groups, the dual-task attention group, and the control
group
< MrEmile> 7 is fluid intelligence too, 8 is probably reading (or word
memorisation)
<gwern> wait, 7 is gf too?
< MrEmile> yup
<gwern> MrEmile: what are the 4 groups on table 7? x, control, y, and control?
< MrEmile> dual n back
< MrEmile> (first column is num. days)
<gwern> x and y are both dual n-back, but then why are the first
characters different? in the second column
< MrEmile> they aren't, the top of the first char of x is just badly printed
<gwern> oh
< MrEmile> I don't know about the N though
< MrEmile> I couldn't find a section called "experimental setup"
< MrEmile> and gotta get away from the computer for a bit now
<gwern> ok, thanks
< MrEmile> ah wait
< MrEmile> section 5.2 mentions 74 kids in fifth grade, 39 boys, 35 girls
<gwern> 5.2, so I guess that would apply only to the second table
< MrEmile> section 4.2 says fifth grade too, 283: 146 boys, 137 girls
<gwern> wow. good sample sizes
<gwern> not what one expects from a master's thesis in the states, anyway
< MrEmile> The Chinese have more raw material to work with. They do
have a bit of a reputation for academic fraud too though.
<gwern> MrEmile: yeah. I'm actually in the middle of trying to get the
raw data for Qiu et al 2009 and all the authors are ignoring me. I
can't tell whether it's their email addresses are dead (no bounces
though) or their weird graphs do in fact indicate some fraud or
sloppiness
<gwern> MrEmile: but neither of those are evenly divisible by 4, so
there must be some breakdown into the experimentals and control
group...
< MrEmile> Section 3.2.1 says 48 fifth-graders, 26 boys, 22 girls,
average age 10.35
If no Chinese-readers here feel like volunteering to check
http://www.doc88.com/p-397166703921.html for details, I'll probably
wind up entering this study into the meta-analysis as 2 studies; I'll
assume the 4 groups were evenly divided, and in each, I will pool the
single group with the dual n-back group and the dual-attention task
group with the control group. He didn't mention anything that
indicated the control group was not no-contact, so that stays the
same.
--
gwern
http://www.gwern.net
Michael
Jun 30, 2012, 12:22:49 PM6/30/12
to brain-t...@googlegroups.com
lol, that conversation sounds analogous to a very tedious math problem. Nice efforts Gwern.
Gwern Branwen
Feb 18, 2013, 9:02:54 PM2/18/13
to brain-t...@googlegroups.com, Emile Kroeger
I've done some more work on this:
- original encrypted file: http://dl.dropbox.com/u/85192141/2012-zhong.ebt (8M)
- screenshots of all pages in thesis:
http://dl.dropbox.com/u/85192141/2011-zhong.tar (20M)
With these copies, hopefully the thesis will still be accessible when
http://www.doc88.com/p-397166703921.html inevitably dies or goes
offline. Which doesn't help with understanding the paper's data,
unfortunately.
Here's my current understanding of the paper and areas where I run
into confusion. If anyone in the group knows Chinese, please help me
out! For my meta-analysis, I need to know each control and training
group, their exact size, their mean IQ score before and after
training, and the standard deviation of their mean.
# experiment 1
> We have trained and control subjects took a pre-test and post-test of 1) fluid intelligence, 2) working memory reading span, and 3) working memory digit span before and after the training.
>
> The trained subjects accepted either working memory training or attention training, only one task each, and meanwhile, the control group didn't do anything. It should be pointed out that there is difference in the amount of training between the trained groups. There are four training levels (five days, ten days, fifteen days, twenty days) of training amount.
multiple levels of training;
sample size? < MrEmile> Section 3.2.1 says 48 fifth-graders, 26 boys,
1. single task of visual working memory
2. working memory dual n-back
3. dual-channel attention task
4. control task
(and the same for 10, 15 and 20-days)
But in table 4, the first line is a mystery: it seems to have the same
label as the last entry for the 5-day groups, which should mean it's
the 'control task'. Except the same characters are repeated in the
10-day and 15-day groups - and not the 20-day group! The 20-day group
actually has just 3 entries, apparently omitting the control group.
What on earth does this mean? There was a 0-day control subgroup and
no control subgroup for the 20-day group?
Also, there seem to be too many groups. If there are 4 different
subgroups for the 5, 10, 15, and 20-day groups, then that's 4x4 or 16
subgroup in total, implying 3 kids in each subgroup (48 / 4*4). The
standard deviations seem too small for each group to be so skimpy.
Section 3's sample size of 48 fifth-graders may refer to the
preliminary experiment (notice that while the abstract talks about
there being 3 experiments, it only really discusses 2), but if we go
with 4.2's claim of 283 children - much more reasonable to be
randomized into 16 subgroups - then we don't get an even division (283
/ 16 = 17.6875).
Another difficulty is that I don't know what IQ test is being used to
measure fluid intelligence here. I checked the list of English
references but there's nothing I recognized as an IQ test; at the end
there's an appendix giving example of what look like geometric
pattern-matching problems... they're very easy but I'm not sure that
they're actual matrix IQ tests, since they could be something else.
So while table 4 on pg 24/18 seems to give the pre/post means and SDs,
I remain uncertain what each line is and what _n_ they represent.
# experiment 2
> The goal of experiment 2 is to explore whether 15 days of training has a special role on improving fluid intelligence; The benefit of low group on fluid intelligence, reading span, digital span is greater than the high group. Experiment 2 hypothesis: Fifteen days of training have special effect to improve fluid intelligence; The low group get benefit more than the high group. The first independent variable includes 15 days and 20 days two level; The second independent variable includes training, control two level. The other conditions are the same as experiment 1.
< MrEmile> section 5.2 mentions 74 kids in fifth grade, 39 boys, 35 girls
<gwern> 5.2, so I guess that would apply only to the second table
pg30/24
Table 7, mercifully, seems easier to interpret. The first 2 rows are
for the 15-day training, the last 2 rows are for 20-days; the first
row is the 15-day single n-back, second row control, third row 20-day
single n-back, fourth row 20-day control group.
The problem here is both that the sample size doesn't seem right since
it doesn't divide evenly into the claimed number of groups (74/4=18.5)
and also the scores are too high: table 4 had scores ~5, while table 7
has scores ~20. Both experiments are supposed to be using fifth
graders, so it can't be that the kids are just older and scoring
higher... Is table 7 actually fluid intelligence/IQ scores? (One might
wonder about table 8, but that has the opposite problem: the average
score is ~2.5!)
Table 9 on pg31/25 has scores in the right range (~5), the training
groups improved more than the controls, and also the 15-day group
scored a higher average than the 19-day group on the post-test - but
Emile didn't mention anything about table 9, so I don't know if that
is the before/after IQ scores for experiment 2.
--
gwern
http://www.gwern.net/DNB%20FAQ#meta-analysis
- original encrypted file: http://dl.dropbox.com/u/85192141/2012-zhong.ebt (8M)
- screenshots of all pages in thesis:
http://dl.dropbox.com/u/85192141/2011-zhong.tar (20M)
With these copies, hopefully the thesis will still be accessible when
http://www.doc88.com/p-397166703921.html inevitably dies or goes
offline. Which doesn't help with understanding the paper's data,
unfortunately.
Here's my current understanding of the paper and areas where I run
into confusion. If anyone in the group knows Chinese, please help me
out! For my meta-analysis, I need to know each control and training
group, their exact size, their mean IQ score before and after
training, and the standard deviation of their mean.
# experiment 1
> We have trained and control subjects took a pre-test and post-test of 1) fluid intelligence, 2) working memory reading span, and 3) working memory digit span before and after the training.
>
> The trained subjects accepted either working memory training or attention training, only one task each, and meanwhile, the control group didn't do anything. It should be pointed out that there is difference in the amount of training between the trained groups. There are four training levels (five days, ten days, fifteen days, twenty days) of training amount.
>
> There are five major research findings:
> (1) The training gain was well explained by a linear function for every group of working memory and attention training tasks
> There are five major research findings:
> (1) The training gain was well explained by a linear function for every group of working memory and attention training tasks
> (2) Fifteen days of training is best to improve fluid intelligence and working memory reading span
> (3) The dual n-back working memory task is the best task to increase fluid intelligence
> (4) The ANOVA and Covariance ANOVA found that participants with initially lower fluid intelligence, working memory reading and digit span showed even larger gains on them
> (5) Training is good for improving math scores.
multiple levels of training;
sample size? < MrEmile> Section 3.2.1 says 48 fifth-graders, 26 boys,
22 girls, average age 10.35
< MrEmile> section 4.2 says fifth grade too, 283: 146 boys, 137 girls
on page 24 [pg18], 4.5.4 table 4 the lines are for 5-days:
1. single task of visual working memory
2. working memory dual n-back
3. dual-channel attention task
4. control task
(and the same for 10, 15 and 20-days)
But in table 4, the first line is a mystery: it seems to have the same
label as the last entry for the 5-day groups, which should mean it's
the 'control task'. Except the same characters are repeated in the
10-day and 15-day groups - and not the 20-day group! The 20-day group
actually has just 3 entries, apparently omitting the control group.
What on earth does this mean? There was a 0-day control subgroup and
no control subgroup for the 20-day group?
Also, there seem to be too many groups. If there are 4 different
subgroups for the 5, 10, 15, and 20-day groups, then that's 4x4 or 16
subgroup in total, implying 3 kids in each subgroup (48 / 4*4). The
standard deviations seem too small for each group to be so skimpy.
Section 3's sample size of 48 fifth-graders may refer to the
preliminary experiment (notice that while the abstract talks about
there being 3 experiments, it only really discusses 2), but if we go
with 4.2's claim of 283 children - much more reasonable to be
randomized into 16 subgroups - then we don't get an even division (283
/ 16 = 17.6875).
Another difficulty is that I don't know what IQ test is being used to
measure fluid intelligence here. I checked the list of English
references but there's nothing I recognized as an IQ test; at the end
there's an appendix giving example of what look like geometric
pattern-matching problems... they're very easy but I'm not sure that
they're actual matrix IQ tests, since they could be something else.
So while table 4 on pg 24/18 seems to give the pre/post means and SDs,
I remain uncertain what each line is and what _n_ they represent.
# experiment 2
> The goal of experiment 2 is to explore whether 15 days of training has a special role on improving fluid intelligence; The benefit of low group on fluid intelligence, reading span, digital span is greater than the high group. Experiment 2 hypothesis: Fifteen days of training have special effect to improve fluid intelligence; The low group get benefit more than the high group. The first independent variable includes 15 days and 20 days two level; The second independent variable includes training, control two level. The other conditions are the same as experiment 1.
< MrEmile> section 5.2 mentions 74 kids in fifth grade, 39 boys, 35 girls
<gwern> 5.2, so I guess that would apply only to the second table
<gwern> MrEmile: what are the 4 groups on table 7? x, control, y, and control?
< MrEmile> dual n back; first column is num. days
Table 7, mercifully, seems easier to interpret. The first 2 rows are
for the 15-day training, the last 2 rows are for 20-days; the first
row is the 15-day single n-back, second row control, third row 20-day
single n-back, fourth row 20-day control group.
The problem here is both that the sample size doesn't seem right since
it doesn't divide evenly into the claimed number of groups (74/4=18.5)
and also the scores are too high: table 4 had scores ~5, while table 7
has scores ~20. Both experiments are supposed to be using fifth
graders, so it can't be that the kids are just older and scoring
higher... Is table 7 actually fluid intelligence/IQ scores? (One might
wonder about table 8, but that has the opposite problem: the average
score is ~2.5!)
Table 9 on pg31/25 has scores in the right range (~5), the training
groups improved more than the controls, and also the 15-day group
scored a higher average than the 19-day group on the post-test - but
Emile didn't mention anything about table 9, so I don't know if that
is the before/after IQ scores for experiment 2.
--
gwern
http://www.gwern.net/DNB%20FAQ#meta-analysis
Gwern Branwen
Feb 19, 2013, 1:47:52 PM2/19/13
to Emile Kroeger, brain-t...@googlegroups.com
On Tue, Feb 19, 2013 at 5:03 AM, Emile Kroeger <flam...@gmail.com> wrote:
> Nope - the lines are grouped by four, but on the leftmost colon the number
> of days is written on the *second line* of the group of four lines. So the
> first line is the first line of the 5-day groups, the second line (on which
> "5 days") is written is actually the second line of the 5-day group, etc.
>
> (I guess he wanted to put it in the middle but screwed up)
Wow, that's awful typography then. Why would you put it in the second
line... or even the middle, if we assume they screwed up.
I hope the answer to the question of 'how many subjects were in each
group' doesn't turn out to be similarly irritating.
--
gwern
http://www.gwern.net
> Nope - the lines are grouped by four, but on the leftmost colon the number
> of days is written on the *second line* of the group of four lines. So the
> first line is the first line of the 5-day groups, the second line (on which
> "5 days") is written is actually the second line of the 5-day group, etc.
>
> (I guess he wanted to put it in the middle but screwed up)
Wow, that's awful typography then. Why would you put it in the second
line... or even the middle, if we assume they screwed up.
I hope the answer to the question of 'how many subjects were in each
group' doesn't turn out to be similarly irritating.
--
gwern
http://www.gwern.net
Reply all
Reply to author
Forward
0 new messages