I suggest that you take a look at the Reaction Time (RT) Color experiment
at the Online Psychological Laboratory (OPL) at http://opl.apa.org .
If you want to collect data from your class, register your class on the
website (see "Register" on the menu bar near the top of the page).
Your class' data will then become publicly available via the Data
page on the website (see "Data" on the menu bar).  If you just want
data that other classes have contributed, one can just go to the data
page and decide how much data one wants to download -- I suggest
a couple of hundred participants for reasons I provide below.

For background on the RT-Color experiment see:
http://opl.apa.org/Experiments/About/AboutReactionTimeColor.aspx
However, the three RT tasks that are used were originally developed
by Franciscus Donders back in in the 1850s and published at time
in, I believe, French and an English translation was published in 1969
in the journal Acta Psychologica (in Pubmed.gov, this is the citation
http://www.ncbi.nlm.nih.gov/pubmed/5811531 ).
The RT tasks are:
(1) Simple RT: press a key on a keyboard when a colored square
appears on the screen
(2) Go-No Go RT: press a key when a red square appears, withhold
response for any other color
(3) Choice RT: press one key for red, another key for blue
The dependent variable is RT in seconds which should be converted
into milliseconds.  There are a minimum of 20 trials for each task
and the mean RT across the 20 trials is used to reflect a person's
performance on the task.  Thus, we have a one-way, three level
within-subject design.  It is expected that mean RT (across subjects)
will increase as a function of task complexity, so:
Mean RT(simple) < mean RT(Go-No Go) < mean RT(choice)

Now, for tasks (2) and (3) error trials can easily be specified, so 20 trials
of correct responses are required (that is, a person may only experience
20 trials on a task if the responses are all correct, more trials if errors are
made).  The Simple RT task doesn't have an "error responses", 
probably because it is assumed that participants should always be able
to do such a task without error.  However, there are some people who
have unusually long mean simple RT that simply can't be right.  It is unclear
what is going on but some people just have a hard time with the simple RT
task (Note: according to Donders' theory and most results, a simple RT
should not be greater than the Go-No Go RT or the Choice RT).  This
is a good discussion point about what do with outliers or extreme values
that have a big influence on the mean.  Median  RT and the minimum RT
values typically follow the predicted results as do trimmed means and
other robust measures of location (such as those used in SPSS Examine
procedure). 

One way to deal with this problem is to exclude subjects/participants 
with simple RT > 1000 ms but one should make sure that they understand
the patterns of RT across all tasks (there usually are outliers there as well;
what do about them -- just ignore them?).  With small samples, like with
a small class, the ANOVA results and multiple comparisons may or may
not be significant because of a lack of statistical power.  Excluding subjects
will make the situation worse.  Getting additional subjects will provide
sufficient power producing a significant repeated measures ANOVA and
multiple comparison results (i.e., all three task means should be significantly
different from each other).

A better solution to the situation is alter the simple RT task a little, say,
if a person's RT on a trial is greater than 1 second, they have to repeat
the trial.  This would then produce 20 "error free" trials comparable to
the other tasks.

-Mike Palij
New York University
[email protected]


------------------------ Original Message
On Sun, 06 Mar 2011 10:08:48 -0800, Michelle Everson wrote:
>Hello,
>First, I apologize for the cross-posting.
>
>I'm looking for a quick in-class activity I can do (in an intermediate
>statistics course) in order to introduce my students to the repeated-
>measures (or one-way within-subjects) ANOVA. I have a relatively small class
>(21 students) and we meet in a computer lab. I was hoping to do something in
>class that would yield some data that we can then explore and analyze right
>on the spot. Does anyone have any good suggestions?  One thought I had was
>something where students could work through puzzles of some kind on the
>computer (where the DV would be the amount of time taken to complete the
>puzzle).  Ideally, I wanted at least three different conditions for them to
>work through, but we could have just two.  I'm just not sure if there is a
>website out there that might give me what I need to pull this off (by
>Wednesday!).  Also, I have some students in class for whom English is a
>second language, so I want to avoid tasks that could be very challenging or
>frustrating for those students (e.g., I was initially thinking of word
>puzzle activities like Boggle, but I think that would be too much for some
>students).
>
>Thank you so much for any ideas you might have to share.

---
You are currently subscribed to tips as: [email protected].
To unsubscribe click here: 
http://fsulist.frostburg.edu/u?id=13090.68da6e6e5325aa33287ff385b70df5d5&n=T&l=tips&o=9208
or send a blank email to 
leave-9208-13090.68da6e6e5325aa33287ff385b70df...@fsulist.frostburg.edu

Reply via email to