how do i get group x time contrasts for 3 groups

[Stuart Cathcart]

Hi list,

Firstly, Im not sure how to use this list so hopefully dont breach any
protocols. I dont know how to see my message or if i have had any
responses, so any help on that basic skill would be appreciated.

Now to my question. I have data measured over time in 3 groups and want to
compare the 3 groups on time effects. I have done a repeated measures anova
and got group and time effects and a group x time interaction, as expected.
i dont know how to do group x time contrasts for particular groups (i.e. to
see which groups differed in time effects). post hoc tests seem to test for
group differences, but not groupxtime differences. any help greatly
appreciated. Thanks.

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Zdaniuk, Bozena-2]

Stuart, here is the syntax that should do what you want to do. Basically, you run your RMANOVA using spss GLM syntax (also available through window menu) and the /EMMEANS subcommand compares all combinations of cells.
Syntax below
GLM
  Time1 time2 time3 time4  BY group
  /WSFACTOR = time  4 Polynomial
  /METHOD = SSTYPE(3)
  /PLOT = PROFILE(time*group)
  /EMMEANS = TABLES(group*time) COMPARE (time)
  /EMMEANS = TABLES(group*time) COMPARE (group)
  /PRINT = DESCRIPTIVE
  /CRITERIA = ALPHA(.05)
  /WSDESIGN = time
  /DESIGN = group .

Let me know if that works and/or if you have any further questions



Bozena Zdaniuk, Ph.D.
University of Pittsburgh
UCSUR, 6th Fl.
121 University Place
Pittsburgh, PA 15260
Ph.: 412-624-5736
Fax: [phone]
Email: [hidden email]<mailto:[hidden email]>

From: stuart cathcart [mailto:[hidden email]]
Sent: Wednesday, August 27, 2008 8:40 PM
To: Zdaniuk, Bozena
Subject: FW: how do i get group x time contrasts for 3 groups


Hi Bozena,

Thanks for your response,

I got the following response and exchange, i guess it's a way to do it (post-hoc t-tests on change scores, and do bonferroni correction i'd assume), but it's not as elegant as i woud have hoped.

to be clear on what im trying to do: we have 3 groups measured over 4 time points. we expect groups 1 and 2 to decrease but not group 3, and, we expect group1 to decrease more than group 2. We did RMANOVA which showed G, T and GxT effect, so that partially supports the hypothesis. But what we dont know where the differences are. so what we need is to do some kind of 'contrasts' to see if a) are groups 1 and 2 decreasing but not group 3, and b) is group 1 decreasing more than group 2.

make sense?

cheers
stuart

________________________________

Date: Wed, 27 Aug 2008 09:34:03 -0400
From: [hidden email]
Subject: Re: how do i get group x time contrasts for 3 groups
To: [hidden email]

Hi Stuart......
    The simple effects are what you are testing for the G x T interactions. The rmanova tells you there are differences between the 6 means, but not where those differences lie.

    If you want to examine rates of change, you could always compute difference scores between time 1 & 2, time 2 & 3 and time 1 & 3 then compare those difference scores between group in an independent samples t-test.

Pete

stuart cathcart wrote:


Hi Pete,

Thanks for getting back to me.

Im not sure what you mean by simple effects for G x T main effects.

using my below email as an exmple, for the interaction, if i do the split file paired t-test, i would see that, for example, groups 1 and 2 both decresed significantly over time while group 3 did not. But, what i really want, is to see if group 1 decreased MORE than group 2 did, so i need somthing like a comparison of 'the rate of change' between, say, group 1 compared to groups 2 and 3.

cheers
stuart



________________________________

Date: Tue, 26 Aug 2008 09:45:52 -0400
From: [hidden email]<mailto:[hidden email]>
Subject: Re: how do i get group x time contrasts for 3 groups
To: [hidden email]<mailto:[hidden email]>

Hi Stuart,
    Let me make sure I have the proper understanding of what you want to test. You are interested in running contrasts on the interaction? The simple effects should be easy enough to run for the G & T main effects, but the simple effects for the interaction can be handled a number of ways, but it depends on what hypothesis you're testing.
    The simplest way would be to split the file by group and then run paired samples tests between the levels of Time. Of course, if you obtain descriptives, you could always just calculate it by hand, which is very useful if you want to run Newman-Keuls or Tukey tests.

Am I on the right track?

Pete

stuart cathcart wrote:


Hi Pete,

Not adequately. Someone responded explaining how to do the basic RMANOVA with 2 groups, (G, T and GxT) but im after how to compare 3 groups on these effects. i.e we're expecting all groups' scores to decrease over time, but are expecting group 1 scores to decrease over time more than gorups 2 and 3. doing the basic RMANOVA, we have got a G, T and GxT effect as expected, but i dont know how to see if group 1 is changing over time more than the other groups.

Cheers
Stuart

> Date: Mon, 25 Aug 2008 21:45:00 -0400
> From: [hidden email]<mailto:[hidden email]>
> To: [hidden email]<mailto:[hidden email]>
> Subject: Re: how do i get group x time contrasts for 3 groups
>
> Stuart,
> Did anyone address this for you?
>
> Pete
>
> Stuart Cathcart wrote:
> > Hi list,
> >
> > Firstly, Im not sure how to use this list so hopefully dont breach any
> > protocols. I dont know how to see my message or if i have had any
> > responses, so any help on that basic skill would be appreciated.
> >
> > Now to my question. I have data measured over time in 3 groups and want to
> > compare the 3 groups on time effects. I have done a repeated measures anova
> > and got group and time effects and a group x time interaction, as expected.
> > i dont know how to do group x time contrasts for particular groups (i.e. to
> > see which groups differed in time effects). post hoc tests seem to test for
> > group differences, but not groupxtime differences. any help greatly
> > appreciated. Thanks.
> >
> > =====================
> > To manage your subscription to SPSSX-L, send a message to
> > [hidden email]<mailto:[hidden email]> (not to SPSSX-L), with no body text except the
> > command. To leave the list, send the command
> > SIGNOFF SPSSX-L
> > For a list of commands to manage subscriptions, send the command
> > INFO REFCARD
> >

________________________________
in your area now! View photos of singles<http://a.ninemsn.com.au/b.aspx?URL=http://dating.ninemsn.com.au/channel/index.aspx?trackingid%3D1046247&_t=773166080&_r=WL_TAGLINE&_m=EXT>

________________________________
with mobile Hotmail SMS alerts  Win a Nokia E51 <http://www.livelife.ninemsn.com.au/compIntro.aspx?compId=4589>

________________________________
Sell your car for just $40 at CarPoint.com.au It's simple! <http://a.ninemsn.com.au/b.aspx?URL=http%3A%2F%2Fsecure%2Dau%2Eimrworldwide%2Ecom%2Fcgi%2Dbin%2Fa%2Fci%5F450304%2Fet%5F2%2Fcg%5F801459%2Fpi%5F1004813%2Fai%5F859641&_t=762955845&_r=tig_OCT07&_m=EXT>

====================To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD