post hoc analyses of within-subject variables

Moderators: statman, Analyst Techy, andris, Fierce, GerineL, Smash

student2008
Posts: 5
Joined: Sun Jul 27, 2008 3:10 am

post hoc analyses of within-subject variables

Postby student2008 » Sun Jul 27, 2008 3:31 am

We ran a four-way ANOVA on RT (reaction time) data with 4 independent within-subject variables. The analysis revealed a number of significant main effects and interactions. Among them there was a significant three-way interaction of three IVs: A (before vs. after), B (type 1 vs. type 2), & C (short vs. long).

We would like to run separate post hoc analyses. Purpose: to see if the interaction of A & B was only significant for one level of C (short or long).

Please advise, we are using version 13 of SPSS.

Many many thanks in advance. :wink:
statman
Administrator
Posts: 2699
Joined: Tue Jun 12, 2007 12:08 pm
Location: Florida, USA

Postby statman » Sun Jul 27, 2008 1:28 pm

Actually, aren't these 2-way interactions?

Anyway, post hoc needs at least 3 variables in the interaction but if they are actually 2-way, as I think, then do a separate (paired) ANOVA or t-tests or even perhaps a chi-sq
See the note below

NOTE: Please read the Posting Guidelines and always tell us your OS, the SPSS version and information about your study and data!

Statman
Statistical Services
student2008
Posts: 5
Joined: Sun Jul 27, 2008 3:10 am

Postby student2008 » Mon Jul 28, 2008 6:39 am

Hi Statman,

Many thanks for your reply. :D
May I ask for further advice?
We are pretty sure about the significant difference between the two levels of C. We would like to explore if the interaction of A & B was only significant for a particular level of C (short or long).
We split the original file into two, according to the two levels of C; then for each separate file we ran GLM repeated measures with A & B as independent within-subject variables.
Not sure if this way of doing it is proper.
Please advise.
Many thanks in advance!
statman
Administrator
Posts: 2699
Joined: Tue Jun 12, 2007 12:08 pm
Location: Florida, USA

Postby statman » Mon Jul 28, 2008 12:13 pm

Might look into an independent t-test with the grouping variable as C and the IVs your A & B. This should do it but also realized that no indication of IV variable type: scale, nominal...

Who is online

Users browsing this forum: No registered users and 1 guest

cron