Contrasts Using TEST Subcommand in MIXED for 3-Way Interaction Model
Locked
 
|
All,
I recently showed how to construct a couple contrasts using the TEST subcommand in MIXED for a 3-way interaction model: http://www.listserv.uga.edu/cgi-bin/wa?A2=ind1101&L=spssx-l&P=R43996 I took a shortcut by treating one of the dichotomous variables as continuous. Below I show how to employ the TEST subcommand when all three variables are treated as categorical. (Note: I try to present the code in such a way that one can see how the TEST statements build upon each other.) HTH, Ryan -- mixed y by condition country gender /fixed=condition country gender condition*country condition*gender country*gender condition*country*gender | sstype(3) /print=solution lmatrix /test 'cond=0,cntry=1,gender=0' intercept 1 condition 1 0 country 1 0 0 gender 1 0 condition*country 1 0 0 0 0 0 condition*gender 1 0 0 0 country*gender 1 0 0 0 0 0 condition*country*gender 1 0 0 0 0 0 0 0 0 0 0 0 /test 'cond=1,cntry=1,gender=0' intercept 1 condition 0 1 country 1 0 0 gender 1 0 condition*country 0 0 0 1 0 0 condition*gender 0 0 1 0 country*gender 1 0 0 0 0 0 condition*country*gender 0 0 0 0 0 0 1 0 0 0 0 0 /test '(cond=0,cntry=1,gender=0) MINUS (cond=1,cntry=1,gender=0)' intercept 0 condition 1 -1 country 0 0 0 gender 0 0 condition*country 1 0 0 -1 0 0 condition*gender 1 0 -1 0 country*gender 0 0 0 0 0 0 condition*country*gender 1 0 0 0 0 0 -1 0 0 0 0 0 /test 'cond=0,cntry=2,gender=0' intercept 1 condition 1 0 country 0 1 0 gender 1 0 condition*country 0 1 0 0 0 0 condition*gender 1 0 0 0 country*gender 0 0 1 0 0 0 condition*country*gender 0 0 1 0 0 0 0 0 0 0 0 0 /test 'cond=1,cntry=2,gender=0' intercept 1 condition 0 1 country 0 1 0 gender 1 0 condition*country 0 0 0 0 1 0 condition*gender 0 0 1 0 country*gender 0 0 1 0 0 0 condition*country*gender 0 0 0 0 0 0 0 0 1 0 0 0 /test '(cond=0,cntry=2,gender=0) MINUS (cond=1,cntry=2,gender=0)' intercept 0 condition 1 -1 country 0 0 0 gender 0 0 condition*country 0 1 0 0 -1 0 condition*gender 1 0 -1 0 country*gender 0 0 0 0 0 0 condition*country*gender 0 0 1 0 0 0 0 0 -1 0 0 0 /test '(cond=0 minus cond=1|cntry=1,gender=0) MINUS (cond=0 minus cond=1|cntry=2,gender=0)' intercept 0 condition 0 0 country 0 0 0 gender 0 0 condition*country 1 -1 0 -1 1 0 condition*gender 0 0 0 0 country*gender 0 0 0 0 0 0 condition*country*gender 1 0 -1 0 0 0 -1 0 1 0 0 0 /emmeans=tables(condition*country*gender). ===================== 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 |
Linear contrasts for a 2x3 mixed design
|
|
Dear all,
I have a dataset from a 2 (between subject IV) x 3 (within subject IV) mixed design experiment. I would like to look at a contrast comparing the mean from one condition with the average mean from the other 5 conditions. My question is: How do we compute the error term for this contrast? Thank you for your attention. Cheers, Yonghao |
Re: Linear contrasts for a 2x3 mixed design
|
|
Why would you want to do that. It confounds the effects of the between and within subjects effects. Did you have a significant interaction for the between subjects by within subjects effects? Dr. Paul R. Swank, Professor and Director of Research Children's Learning Institute University of Texas Health Science Center-Houston From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Lim Yonghao Dear all, |
Re: Linear contrasts for a 2x3 mixed design
|
|
Dear Prof Swank,
There is a significant interaction. The solution i had was just to look as another interaction and conduct the appropriate posthoc tests. However, i am just wondering whether there is an contrast that enables me to investigate that. Thank you for your attention. Cheers, Yonghao Date: Mon, 31 Jan 2011 09:48:08 -0600 From: [hidden email] Subject: Re: Linear contrasts for a 2x3 mixed design To: [hidden email] Why would you want to do that. It confounds the effects of the between and within subjects effects. Did you have a significant interaction for the between subjects by within subjects effects?
Dr. Paul R. Swank, Professor and Director of Research Children's Learning Institute University of Texas Health Science Center-Houston
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Lim Yonghao
Dear a
ll, |
Re: Linear contrasts for a 2x3 mixed design
|
|
If A has two categories and B has three then Put contrasts on the B variable and consider the interaction of the A variable with each contrast of the B variable. Dr. Paul R. Swank, Professor and Director of Research Children's Learning Institute University of Texas Health Science Center-Houston From: Lim Yonghao [mailto:[hidden email]] Dear Prof Swank, Date: Mon, 31 Jan 2011 09:48:08 -0600 Why would you want to do that. It confounds the effects of the between and within subjects effects. Did you have a significant interaction for the between subjects by within subjects effects? Dr. Paul R. Swank, Professor and Director of Research Children's Learning Institute |
Re: Linear contrasts for a 2x3 mixed design
Administrator
|
In reply to this post by Swank, Paul R
I have another question: What is the within-subjects variable?
If it is something for which polynomial contrasts would be appropriate (i.e., if there is an underlying quantitative variable), you would get a table of contrast results showing tests for Group x Linear and Group x Quadratic terms. Those, along with a decent graph would be sufficient to explain the nature of the interaction for most folks, I think. HTH.
--
Bruce Weaver bweaver@lakeheadu.ca http://sites.google.com/a/lakeheadu.ca/bweaver/ "When all else fails, RTFM." PLEASE NOTE THE FOLLOWING: 1. My Hotmail account is not monitored regularly. To send me an e-mail, please use the address shown above. 2. The SPSSX Discussion forum on Nabble is no longer linked to the SPSSX-L listserv administered by UGA (https://listserv.uga.edu/). |
|
|
In reply to this post by Lim Yonghao
Yonghao,
Your 2X3 design can be shown as: B b1 b2 b3 -------------------------- a1 a1b1 | a1b2 | a1b3 A -------------------------- a2 a2b1 | a2b2 | a2b3 -------------------------- If you wanted to test [using MIXED] whether the a1b1 cell mean is significantly different than the mean of the other 5 cell means, then you would write the following TEST statement: /TEST 'a1b1 minus others' a 3 -3 b 4 -2 -2 a*b 5 -1 -1 -1 -1 -1 divisor=5. Ryan 2011/1/31 Lim Yonghao <[hidden email]>: > Dear all, > > I have a dataset from a 2 (between subject IV) x 3 (within subject IV) mixed > design experiment. I would like to look at a contrast comparing the mean > from one condition with the average mean from the other 5 conditions. > > My question is: How do we compute the error term for this contrast? > > Thank you for your attention. > > Cheers, > Yonghao > ===================== 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 |
Re: Linear contrasts for a 2x3 mixed design
|
|
In reply to this post by Lim Yonghao
Prof Swank,
That is what i am planning to do if all the syntax provided by Ryan does not work. Thank you very much for the advice. Cheers, Yonghao From: [hidden email] To: [hidden email]; [hidden email] Date: Tue, 1 Feb 2011 12:11:37 -0600 Subject: RE: Linear contrasts for a 2x3 mixed design If A has two categories and B has three then Put contrasts on the B variable and consider the interaction of the A variable with each contrast of the B variable.
Dr. Paul R. Swank, Professor and Director of Research Children's Learning Institute University of Texas Health Science Center-Houston
From: Lim Yonghao [mailto:[hidden email]]
Dear Prof Swank, Date: Mon, 31 Jan 2011 09:48:08 -0600 Why would you want to do that. It confounds the effects of the between and within subjects effects. Did you have a significant interaction for the between subjects by within subjects effects?
Dr. Paul R. Swank, Professor and Director of Research Children's Learning Institute |
| Free forum by Nabble | Edit this page |