Dear Prof Donald McLaren
I am sorry to disturb you! I have four discrete variables {factor 1(A,B), factor 2(C,D), factor 3(E,F), and gender(M,F)} and a continuous variable (age), I want to use the GLM to analysis the result of the interaction among factor 1(A,B), factor 2(C,D) and factor 3(E,F) regressing out the effect of gender and age, and how to make the contrast? thank you very much!
look forward your reply!
Bo Xiang
At 2012-11-21 04:07:37,"MCLAREN, Donald" mclaren.donald@gmail.com wrote:
On Tue, Nov 20, 2012 at 12:56 PM, Douglas N Greve greve@nmr.mgh.harvard.edu wrote:
Thanks Donald. Is this the standard way to do this? I had used 8 rows instead of 4 with the difference being that 8 rows gives you an opportunity to look for an effect in males OR females.
Yes. Having 8 rows would tell you if you have an interaction between factor 1 and 2 in either males or females. My 4 rows only tell you if the interaction exists. Technically speaking, one would run the three-way interaction first. If nothing existed then you do the two-way interaction as I suggested. If there is a three-way interaction, then you would use Doug's approach of the interaction in either males or females.
If there is an effect in both
males and females but the effects go in opposite directions, then the 4 row implementation will resolve to 0 (no effect). Or am I misunderstanding something (again:)?
Nope. You are right. If the male and female effects are different, then they could cancel each other out. If you suspect this to be the case, then you should be able to demonstrate a three-way interaction.
thanks! doug
On 11/20/2012 01:50 PM, MCLAREN, Donald wrote:
Bo,
Doug asked me to chime in on your issue. Here are some points that you (and others) will hopefully find useful.
(1) Inferences are two-step process. First, you create and estimate the design matrix. Every column in the design matrix accounts can account for some of the variance in the data. Second, you have contrasts that allow you to infer specific effects. Because the model contains your covariates, you are always controlling for the covariates and by extension any factor/covariate not in the contrast.
(2) Forming contrasts is often the most difficult thing to do. I assume that your three factors (1, 2, and gender) are all between-subject factors. If one of them is a within-subject factor please let me know and disregard the rest of the email. The final F-contrast will have 4 rows (factor 1 levels-1)*(factor 2 levels -1)=(3-1)*(3-1)=2*2=4
The following is an outline for creating contrasts: (a) Start simple - difference between levels of 1 factor (b) Define your null hypothesis: AO=AP=AQ (c) Make it equal to 0: AO-AP=0 AND AP-AQ=0 (d) Repeat for the other levels of the factor... BO-BP=0 AND BP-BQ=0 CO-CP=0 AND CP-CQ=0
(e) Now combine them AO-AP=BO-BP=CO-CP AND AP-AQ=BP-BQ=CP-CQ
(f) Make them equal to 0: AO-AP-BO+BP=0 BO-BP-CO+CP=0 AP-AQ-BP+BQ=0 BP-BQ-CP+CQ=0
(g) Expand them to include gender, for example: AO-AP-BO+BP=0 becomes FAO-FAP-FBO+FBP+MAO-MAP-MBO+MBP=0
Since the contrast now has 2 columns per level, you should divide all values by 2. This will produce the correct amplitude and statistics. If you leave the values as 1 and -1, then you will have an incorrect amplitude, but the statistics will still be correct.
(h) Fill in the respective columns of your design matrix.
(3) The degrees of freedom are defined based on the rows of the F-matrix and the number of rows in the design matrix. The F-test has a numerator and denominator degrees of freedom. F(n,d).
Best Regards, Donald McLaren
D.G. McLaren, Ph.D. Research Fellow, Department of Neurology, Massachusetts General Hospital and Harvard Medical School Postdoctoral Research Fellow, GRECC, Bedford VA Website: http://www.martinos.org/~mclaren Office: (773) 406-2464 ===================== This e-mail contains CONFIDENTIAL INFORMATION which may contain PROTECTED HEALTHCARE INFORMATION and may also be LEGALLY PRIVILEGED and which is intended only for the use of the individual or entity named above. If the reader of the e-mail is not the intended recipient or the employee or agent responsible for delivering it to the intended recipient, you are hereby notified that you are in possession of confidential and privileged information. Any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited and may be unlawful. If you have received this e-mail unintentionally, please immediately notify the sender via telephone at (773) 406-2464 or email.
On Tue, Nov 20, 2012 at 8:17 AM, xiangbo_2010xiangbo_2010@126.com wrote:
Dear doug Thank you for your reply! +AOM -BOM -APM -BPM 0 0 0 0 0 +AOM -BOM 0 0 -AQM +BQM 0 0 0 +AOM 0 -APM 0 0 0 -COM +CPM 0 +AOM 0 0 0 -AQM -COM 0 +CQM there should be use 1 -1 or 0.5 -0.5? whether the -BPM should be change BPM? Thanks
Bo Xiang
At 2012-11-19 07:23:31,"Douglas Greve"greve@nmr.mgh.harvard.edu wrote:
Hi Bo, you can think of the Ftest as a logical 'OR' between the t-test contrasts indicated in each row. Each row is a difference of differences, so
- (A-B)om - (A-B)pm --> Does the difference between A and B differ
between groups O and P for Males? 2. (A-B)om - (A-B)qm 3. (A-C)om - (A-C)pm 4. (A-C)om - (A-C)qm 5. (A-B)of - (A-B)pf --> Does the difference between A and B differ between groups O and P for Females? 6. (A-B)of - (A-B)qf 7. (A-C)of - (A-C)pf 8. (A-C)of - (A-C)pf
I've put together the first 9 columns of the first 4 rows. The last 9 columns are all 0s. For the last for rows, the 0s and below matrix are swapped to give you the same for the females
doug
AOM BOM APM BPM AQM BQM COM CPM CQM
+AOM -BOM -APM -BPM 0 0 0 0 0 +AOM -BOM 0 0 -AQM +BQM 0 0 0 +AOM 0 -APM 0 0 0 -COM +CPM 0&! nbsp;&nb sp;
+AOM 0 0 0 -AQM -COM 0 +CQM
On 11/17/12 9:21 PM, xiangbo_2010 wrote:
Hi Freesurfer experts,
I'm very sorry to bother you, but I am very confused with the following questions:
My experimental design includes three discrete factors: factor 1 with three levels (A,B,C ); factor 2 with three levels (O,P,Q); gender (F, M), and one covariate.
So I can get 18 classes: FAO, FAP,FAQ,FBO,FBP,FBQ,FCO,FCP,FCQ,MAO, MAP,MAQ,MBO,MBP,MBQ,MCO,MCP,MCQ. I want to perform the interaction between factor 1 and factor 2 regressing out the effect of gender and one covariate, but I don't know the rules for setting the contrasts for the F-test. The contrast matrix I used is:
1 1 1 -1 -1 -1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
1 -1 0 1 -1 0 1 -1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
1 1 1 0 0 0 -1 -1 -1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
1 0 -1 1 0 -1 1 0 -1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 1 1 1 -1 -1 -1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 1 -1 0 1 -1 0 1 -1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 1 1 1 0 0 0 -1 -1 -1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 1 0 -1 1 0 -1 1 0 -1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
is it correct?
Any help will be very appreciated.
Best wishes,
Bo Xiang
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
The information in this e-mail is intended only for the person to whom it is addressed. If you believe this e-mail was sent to you in error and the e-mail contains patient information, please contact the Partners Compliance HelpLine at http://www.partners.org/complianceline . If the e-mail was sent to you in error but does not contain patient information, please contact the sender and properly dispose of the e-mail.
-- Douglas N. Greve, Ph.D. MGH-NMR Center greve@nmr.mgh.harvard.edu Phone Number: 617-724-2358 Fax: 617-726-7422
Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html Outgoing: ftp://surfer.nmr.mgh.harvard.edu/transfer/outgoing/flat/greve/
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer