[Amorya]

Guys, I really need help. Just a little hint would be nice - I'm not asking you to do my work for me. But I'm kinda stuck and very confused!

I have a set of data to analyse. It's results of a study into road safety. The experiment was designed by splitting everyone into two groups - control and intervention. They were both tested on a questionnaire. Then six weeks went by. The intervention group was shown an advert, then they were both tested again.

So there were four conditions - each group had a 'before' and an 'after' condition.

This suggests to me I should use a factorial ANOVA - since the study is both a between-subjects and a within-subjects design.

The columns in my SPSS database are: for each participant: group (1=control, 2=intervention), before and after.

I know to run an ANOVA like this, I need to go to General Linear Model in the SPSS menubar, but do I use a univariate or a multivariate one?

Basically the prediction is that I should have a 2x2 grid like so:

Code:
--------------------------------
|             | before | after |
--------------------------------
| control     |    A   |   B   |
|-------------|--------|-------|
| intervention|    C   |   D   |
--------------------------------

If the advert affected the attitudes significantly, cells A, B and C should be identical (give or take) and D should be significantly different.

Any hints on how to run such an ANOVA on SPSS? I will be exceedingly grateful if anyone can help!!!


Amorya

[Amorya]

Bump?

Surely someone's proficient in SPSS... Or even knows the stats so can advise me in an app-generic way?

[SimpleLife]

Originally posted by Amorya:
Bump?

Surely someone's proficient in SPSS... Or even knows the stats so can advise me in an app-generic way?

Spell out clearly your hypothesis then you will understand better what you need to do.

[philm]

You need to provide a bit more detail on your questionnaire data. Does this provide a score for each individual?

If so, I would be tempted to calculate the difference between before and after and allocated this difference score to each individual. You should then check to see if these differences are normally distributed. You can then run an unpaired t-test between the intervention and control groups to invetigate the effect of your intervention.

Otherwise, you would run the univariate model (you only have one variable - questionaire score) and the construct a repeated measures analysis with 'group' as an additional factor.

[BRussell]

Unfortunately you can't do repeated measures ANOVAs in base SPSS. They make you pay another $49,000 (or whatever outrageous price it is) to get Advanced Models. If you do have that, I believe 'Repeated Measures' should appear underneath General Linear Model along with Univariate and Multivariate. If it's not there, you just can't do it.

Instead you should do what philm suggested (except for that crap about checking to see if they're normally distributed - don't ever do that ), and do an unpaired t-test on the difference scores.

[Amorya]

Originally posted by BRussell:
Unfortunately you can't do repeated measures ANOVAs in base SPSS. They make you pay another $49,000 (or whatever outrageous price it is) to get Advanced Models. If you do have that, I believe 'Repeated Measures' should appear underneath General Linear Model along with Univariate and Multivariate. If it's not there, you just can't do it.

Instead you should do what philm suggested (except for that crap about checking to see if they're normally distributed - don't ever do that ), and do an unpaired t-test on the difference scores.

Ah - thanks loads! I found Repeated Measures on a university copy of SPSS. That was what I was looking for all along!

I think I've got it mostly sorted now. Scores seemed to turn up significant anyhow.


Thanks a lot guys - you've been a real help.


Amorya

[SimpleLife]

Originally posted by SimpleLife:
Spell out clearly your hypothesis then you will understand better what you need to do.

Sounds like I did not get it.

[olePigeon]

42.