Hi Emma,
Thank you so much for your help and sorry for bothering you again. Can I just ask the following:
I input 'chemo' (during/after - 2 levels) as factor 1 and then 'quest' (5 levels for five subscales). The output says that there is no significant effect of chemo but there is a significant effect of quest. What exactly does this mean? I want to know if the 5 subscales differed across the two conditions. Is this output telling me this?
Thanks again for your help and your patience!
Kind Regards,
Nicola Davies,
BSc; MSc Comm.; PhD Researcher
Researcher at the University of Oxford
-----Original Message-----
From: Emma Preece [mailto:[log in to unmask]]
Sent: Sun 11/16/2008 17:48
To: Davies, Nicola
Subject: RE: Testing for Difference between Two Time Points
Hi Nicola,
No problem. Don't worry, anova's produce a lot of data in tabular format and it takes a little getting used to but if I'm correct in thinking that your participants completed all of the questionnaires twice and that each of the three questionnaires are different then a repeated measures anova and this form of data entry should be compatible with your design. If not sorry for confusing you! Some text books are also confusing when it comes to anovas....I'd recommend Dancey and Reidy's Statistics Without Maths for Psychology (any edition) as a good one though. I've also attached an SPSS walkthrough my department devised a while ago and some notes on anovas which might be useful (they were written for an other version of SPSS so something's may look different now although most of the functions are still there somewhere!). These cover several forms of anova and it may be that one of the other types is more appropriate.
In the one I described, the 'tests of within subjects effects' output box is the main one to look at and it will give you both tests of difference & interactions (traditionally using the Greenhouse-Geisser line). For example, if you get a significant main effect for questionnaire/session that would imply that there is a significant difference between the questionnaires or sessions respectively. The interaction would be in this table as questionnaire*session. If any of these are significant that would imply that instead of all of the variables resulting in significant effects it might just be a significant difference between their responses during and after treatment for 1 or 2 of the questionnaires only. Hence, it might suggest that while one or two of the questionnaire are useful tools for differentiating their responses across the two time points another is not so it's informative to include any interactions in your analyses from a methodological perspective. You would then need to go on and run post hoc tests to see exactly where this difference. Hope these help...
Emma
Emma Preece
Research Student / Research Assistant
Office: BB 60 (Psychological Sciences)
Tel: 01905 54 2158 / Mob: 07947 190949
Address: Psychological Sciences, University of Worcester, Henwick Grove, Worcester, WR2 6AJ
_____
From: Research of postgraduate psychologists. on behalf of Davies, Nicola
Sent: Sun 16/11/2008 12:19 PM
To: [log in to unmask]
Subject: Re: Testing for Difference between Two Time Points
Hi Emma,
You have been a great help - thank you so much! I have been playing around with your instructions, as well as with my SPSS book, but am still having some difficulty and was wondering if you had any feedback:
1. My book suggests that ANOVA is used to test for interaction as opposed to difference.
2. By placing 'questionnaire' as the factor, am I saying that the questionnaire responses are IV's? I thought they would be DV's in that responses would be different depending on the condition (i.e. during or after chemo).
3. How would testing whether different subscales on the 3 questionnaires change between the two treatment modalities fit into this method of analysis?
Sorry if these are silly questions, but if you can provide some more info on how an ANOVA shows difference, I would be very grateful. It certainly looks easier than carrying out t-tests, if I can get my head around it!
Kind Regards,
Nicola Davies,
BSc; MSc Comm.; PhD Researcher
Researcher at the University of Oxford
-----Original Message-----
From: Emma Preece [mailto:[log in to unmask]]
Sent: Sun 11/16/2008 10:01
To: Davies, Nicola
Subject: RE: Testing for Difference between Two Time Points
Hi Nicola,
Have you considered using a repeated measures ANOVA as opposed to paired sample's t-tests? Running several t-tests can be problematic and lead misleading results when you're conducting multiple comparisons whereas analyses of variance are more robust and more user friendly. If this is appropriate for your research you would enter your data in relatively the same way as for a paired sample t-test although your columns may need to be a little more specific. Then follow the analyses - general linear model - repeated measures procedure in SPSS. Enter something along the lines of 'questionnaire' in the factor name box and 3 in the number of levels box, click add. Next, you'd enter something like 'session' with 2 levels (i.e. during/after treatment) and click add followed by 'define'. You might need to add more factors if your questionnaires assess more than 1 variable though, so it depends on what you're assessing.
When you have clicked define, you'll get the box where you pair up your variables and it tells you the order to use...so 1,1 would be the questionnaire 1/during treatment data, 1,2 would be questionnaire 1/after treatment data and so on. Then just select posthoc tests and options as appropriate for your design and click ok.
Hope some of this helps!
Emma
Emma Preece
Psychology Research Student / Research Assistant
Office: Bredon Building Room 60
Tel: 01905 54 2158
Address: Psychological Sciences, University of Worcester, Henwick Grove, Worcester, WR2 6AJ.
_____
From: Research of postgraduate psychologists. on behalf of Davies, Nicola
Sent: Sat 15/11/2008 10:34 PM
To: [log in to unmask]
Subject: Testing for Difference between Two Time Points
Hi there,
I wonder if anyone could help me.
I have 35 participants and each participants who completed 3 questionnaires during chemotherapy and the same three questionnaires after chemotherapy. The results of the questionnaires are inputted into SPSS as scale data (calculated from adding up Likert-scale responses). In order to test for a difference between questionnaire responses during and after chemo, would I carry out a PAIRED T-TEST?
Thank you for your help.
Kind Regards,
Nicola Davies,
BSc; MSc Comm.; PhD Researcher
Researcher at the University of Oxford
|