Re: repeated measures in SPSS

On Jan 18, 2:35 pm, Ryan <Ryan.Andrew.Bl...@xxxxxxxxx> wrote:
On Jan 17, 10:04 pm, matt.lew...@xxxxxxxxx wrote:





5 rounds refers to repetitions.

So the design is that everyone involved does five presentations of the
task.

I want to see whether there is improvement on the task over the
presentations, and whether there is a significant effect of
presentation. SO this is the within subject effect

I also want to see if this effect is different between males and
females, so have introduced gender as a between subjects factor.

On Jan 18, 12:52 pm, Richard Ulrich <Rich.Ulr...@xxxxxxxxxxx> wrote:

On Thu, 17 Jan 2008 16:47:30 -0800 (PST), matt.lew...@xxxxxxxxx wrote:

I have a study with 5 rounds which I analyse using Repeated Measures
in SPSS and all comes out with the main effect being significant.

Does "5 rounds" mean 5 periods/repetitions?

I think you are not describing the design sufficiently,
because I think you are not talking about the Within-subject
effect.   Is that the between-subject factor?

BUT when I add in gender as a between subject factor the main effect
isn't significant anymore, and then the interaction is not significant
either which I expected.  I am using a Full Factorial model.

Are you able to tell me why the main effect is not significant
anymore?  I always thought that gender gets added into the model as
the second step rather than upfront.

Well, generically, No.   Unless you select specific options,
the analysis with two Between factors  will be "simultaneous",
one way or another.

Do I have to change it from full factorial to a custom model for this?

Many thanks in advance for your help

You might post the question to the SPSS group,
  comp.soft-sys.stat.spss   along with your command lines.

--
Rich Ulrichhttp://www.pitt.edu/~wpilib/index.html-Hide quoted text -

- Show quoted text -

Are you saying that participants are given a presentation and then
asked to complete a task, and are then given a different presentation
and asked to complete the same task...up to five presentations? Are
these presentations spread out in equal intervals?

It sounds like you want to know if at least two task means are
significantly different from each other, but you are concerned about/
interested in possible carry over effects?

There are problems with running a repeated measures design (without
counterbalancing) when you have carry over effects. How would you
separate presentation effect from carry over effect? There might be a
way, but I don't know of it.

If there is a potential for carry over effects, it would be helpful to
set up a B/N Ss ANOVA (each subject assigned only one presentation) or
a counterbalancing design (change presentation order for each
subject). If you think the effect on task performance for one
presentation affects performances for other presentations differently
(differential carry over effects), then I'd avoid the counterbalancing
approach and stick with a B/N Ss ANOVA design. In other words, if you
had a counterbalanced design such as:

A B

B A

but you thought the carry over effect that A had on the task
performance of B is different than the carry over effect B would have
on the task performance of A, then this approach is problematic.

----
Wish I could be of more help you with your current design.- Hide quoted text -

- Show quoted text -

I reread your post and think I slightly misunderstood. You mean
"presentation of the task" as in each participant is presented the
task (and their performance on the task is scored) five times. I
think my response still holds. There are carry over effects, and thus
"presentation effect" AND the carry over effect will impact
performance of the task. I think this issue needs to be addressed
before worrying about adding other variables to the model.
.