Got it! I guess we may work around by specifying weighted contrast like [ 0.1   0.2   0.3   0.4 ] ?
-- 
Daniel (Yung-Jui) Yang, Ph.D.
Postdoctoral Researcher
Yale Child Study Center
New Haven, CT
Tel: (203) 737-5454
E-mail: yung-jui.yang@yale.edu

On 2/18/14 11:25 PM, "Douglas Greve" <greve@nmr.mgh.harvard.edu> wrote:


Let's say you have four groups with 10 20 30 40 (100 total). The first group would get a weight of .25 where it only had 10% of the total. I don't think there is any way around this.


On 2/18/14 4:07 PM, Yang, Daniel wrote:
Doug, sorry, I am not sure I can fully understand. 

Do you mean the results are weighted MORE toward groups with smaller sample sizes (than groups with larger sample sizes)?

-- 
Daniel (Yung-Jui) Yang, Ph.D.
Postdoctoral Researcher
Yale Child Study Center
New Haven, CT
Tel: (203) 737-5454

On 2/18/14 2:13 PM, "Douglas N Greve" <greve@nmr.mgh.harvard.edu> wrote:

  It will tend to weight smaller groups by a proportion greater than the
number of subjects in the group.



On 02/18/2014 01:58 PM, Yang, Daniel wrote:
Thanks Doug! One follow-up question: since it's [0.25  0.25  0.25
  0.25], would the effect be biased toward large sample sizes in one of
the groups (versus the other groups)?

Or, does FreeSurfer qdec take care of the unequal cell, in a possible
way analogous to using Type III SS in the traditional ANOVA?

Best,
Daniel
--
Daniel (Yung-Jui) Yang, Ph.D.
Postdoctoral Researcher
Yale Child Study Center
New Haven, CT
Tel: (203) 737-5454

On 2/18/14 1:30 PM, "Douglas N Greve" <greve@nmr.mgh.harvard.edu

     That looks right to me
     doug


     On 02/18/2014 10:08 AM, Yang, Daniel wrote:

         Dear FreeSurfer experts,

         I have two binary factors (F1: F1L1 F1L2; F2: F2L1 F2L2), 1
         covariate
         (COV), and 1 nuisance factor (NUI). I want to make sense of some
         specification of the regressors and contrasts.
         Because they are not described
         understanding below. Can you let me know if it is correct?

         This model using the DODS method will have 12 regressors:

         Regressor1: 1's for F1L1 F2L1, 0 otherwise (i.e., offset for
         Group 1)
         Regressor2: 1's for F1L1 F2L2, 0 otherwise (i.e, offset for
         Group 2)
         Regressor3: 1's for F1L2 F2L1, 0 otherwise (i.e., offset for
         Group 3)
         Regressor4: 1's for F1L2 F2L2, 0 otherwise (i.e., offset for
         Group 4)
         Regressor5: COV for Group1 subjects, 0 otherwise (i.e., COV
         slope for
         Group 1)
         Regressor6: COV for Group2 subjects, 0 otherwise (i.e., COV
         slope for
         Group 2)
         Regressor7: COV for Group3 subjects, 0 otherwise (i.e., COV
         slope for
         Group 3)
         Regressor8: COV for Group4 subjects, 0 otherwise (i.e., COV
         slope for
         Group 4)
         Regressor9: NUI for Group1 subjects, 0 otherwise (i.e., NUI
         slope for
         Group 1)
         Regressor10: NUI for Group2 subjects, 0 otherwise (i.e., NUI
         slope for
         Group 2)
         Regressor11: NUI for Group3 subjects, 0 otherwise (i.e., NUI
         slope for
         Group 3)
         Regressor12: NUI for Group4 subjects, 0 otherwise (i.e., NUI
         slope for
         Group 4)

         Description1: Does the correlation between thickness and COV,
         accounting for F1 and F2, differ from 0? Nuisance factors: NUI
         Contrast1: [0  0  0  0  0.25  0.25  0.25  0.25  0  0  0  0]

         Best,
         Daniel

         --
         Daniel (Yung-Jui) Yang, Ph.D.
         Postdoctoral Researcher
         Yale Child Study Center
         New Haven, CT
         Tel: (203) 737-5454


         _______________________________________________
         Freesurfer mailing list


     --
     Douglas N. Greve, Ph.D.
     MGH-NMR Center
     Phone Number: 617-724-2358
     Fax: 617-726-7422

     Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting
     Outgoing:

     _______________________________________________
     Freesurfer mailing list


     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
Phone Number: 617-724-2358
Fax: 617-726-7422

Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting

_______________________________________________
Freesurfer mailing list