Thanks for the quick reply!

So if I understand correctly, the power of say the controls vs. diabetes (indeed it is type 2 diabetes) comparison is constrained by the sample size of the obese group?

2016-10-13 11:02 GMT-03:00 Bruce Fischl fischl@nmr.mgh.harvard.edu:

Hi Eelco

it isn't really a question of whether our implementation is senstitive to this. It's that in general your power will be constrained by the size of the smaller group (I assume this is Type 2 diabetes by the way).

cheers Bruce

On Thu, 13 Oct 2016, Eelco van Duinkerken wrote:

...

Hi all, I am using FS with data from 2 different studies that were acquired on

the

...

same MRI-machine with the same T1 and FLAIR sequences. Unfortunately, the group sizes are not very balanced, with 31 controls, 16 obese and 32 diabetes patients.

Is the GLM for thickness used in FS very sensitive to this unequal group size?

Thanks for the help,

Eelco

---

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.

Ah sorry for the low level of info. I am aiming to do a simple group comparison, where I'm comparing thickness between controls and obese or T2DM patients, regressing out the effect of age, sex and, hypertension. So I created one thickness file with fwhm 10 containing all participants of the 3 groups. The contrast file looks like this:

1 0 -1 0 0 0 for controls vs T2DM 1 -1 0 0 0 0 for controls vs obese 0 1 -1 0 0 0 for obese vs T2DM, although there I do not expect something to happen

And the GLM works fine (differences in the first and second contrast), but I was wondering whether the fact that the groups are unbalanced might influence the results in some way.

Thanks,

Eelco

2016-10-13 11:17 GMT-03:00 Bruce Fischl fischl@nmr.mgh.harvard.edu:

Hi Eelco

it depends on how you setup your GLM. Are you trying to regress out the effects of obesity in some way? If you give us more details I expect someone else can answer your question (Doug!)

cheers Bruce

On Thu, 13 Oct 2016, Eelco van Duinkerken wrote:

Thanks for the quick reply!

...

So if I understand correctly, the power of say the controls vs. diabetes (indeed it is type 2 diabetes) comparison is constrained by the sample size of the obese group?

2016-10-13 11:02 GMT-03:00 Bruce Fischl fischl@nmr.mgh.harvard.edu: Hi Eelco

  it isn't really a question of whether our implementation is
  senstitive to
  this. It's that in general your power will be constrained by the
  size of
  the smaller group (I assume this is Type 2 diabetes by the way).

  cheers
  Bruce


  On Thu, 13 Oct 2016, Eelco van Duinkerken wrote:

  > Hi all,
  > I am using FS with data from 2 different studies that were
  acquired on the
  > same MRI-machine with the same T1 and FLAIR sequences.
  Unfortunately, the
  > group sizes are not very balanced, with 31 controls, 16 obese
  and 32
  > diabetes patients.
  >
  > Is the GLM for thickness used in FS very sensitive to this
  unequal group
  > size?
  >
  >
  > Thanks for the help,
  >
  > Eelco
  >
  >

---

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.

-- Eelco van Duinkerken, PhD Pontifícia Universidade Católica | Department of Psychology | R. Marquês de São Vicente 225 | Gávea, Rio de Janeiro - RJ | CEP 22451-900 | Brasil | E-mail: e.vanduinkerken@vumc.nl | Phone: +55-21-35271855 | http://lattes.cnpq.br/7180895567820901 & VU University Medical Center | Department of Medical Psychology | De Boelelaan 1117 | 1081 HV | Amsterdam | The Netherlands

---

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.

It will be constrained by the group with the smaller number of subjects. There is not anything wrong with unequal group sizes, but it is less efficient than if you had split your acquisitions evenly.

On 10/13/16 10:17 AM, Bruce Fischl wrote:

Hi Eelco

it depends on how you setup your GLM. Are you trying to regress out the effects of obesity in some way? If you give us more details I expect someone else can answer your question (Doug!)

cheers Bruce

On Thu, 13 Oct 2016, Eelco van Duinkerken wrote:

...

Thanks for the quick reply! So if I understand correctly, the power of say the controls vs. diabetes (indeed it is type 2 diabetes) comparison is constrained by the sample size of the obese group?

2016-10-13 11:02 GMT-03:00 Bruce Fischl fischl@nmr.mgh.harvard.edu: Hi Eelco

  it isn't really a question of whether our implementation is
  senstitive to
  this. It's that in general your power will be constrained by the
  size of
  the smaller group (I assume this is Type 2 diabetes by the way).

  cheers
  Bruce


  On Thu, 13 Oct 2016, Eelco van Duinkerken wrote:

  > Hi all,
  > I am using FS with data from 2 different studies that were
  acquired on the
  > same MRI-machine with the same T1 and FLAIR sequences.
  Unfortunately, the
  > group sizes are not very balanced, with 31 controls, 16 obese
  and 32
  > diabetes patients.
  >
  > Is the GLM for thickness used in FS very sensitive to this
  unequal group
  > size?
  >
  >
  > Thanks for the help,
  >
  > Eelco
  >
  >

---

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.

-- Eelco van Duinkerken, PhD Pontifícia Universidade Católica | Department of Psychology | R. Marquês de São Vicente 225 | Gávea, Rio de Janeiro - RJ | CEP 22451-900 | Brasil | E-mail: e.vanduinkerken@vumc.nl | Phone: +55-21-35271855 | http://lattes.cnpq.br/7180895567820901 & VU University Medical Center | Department of Medical Psychology | De Boelelaan 1117 | 1081 HV | Amsterdam | The Netherlands

---

Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer

Thank you Jurgen! That helps a lot. Good to know that the results are not driven by any difference in varience.

Thanks all,

Eelco

2016-10-13 12:00 GMT-03:00 Juergen Haenggi j.haenggi@psychologie.uzh.ch:

Dear Eelco

years ago i posed a similar question on the FS email list

https://mail.nmr.mgh.harvard.edu/pipermail//freesurfer/ 2011-November/021386.html

a part of the response is below

...

...

...

• The other question is how mri_glmfit estimates the variance. Because I have

*>>>* only one subject in the patient group, there is not variance in that group *>>>* and therefore I guess mri_glmfit used pooled (across both groups) estimates *>>>* of variance, isn't it? *>>> >>> >>* Yes, it uses pooled estimates. *>>>>* doug*

if you are concerned with respect to the variance estimates, it does not matter because variance is estimated pooledly, i.e. across both groups. hope this helps

Cheers Jürgen

---

---

University of Zurich Dr. Jürgen Hänggi, Ph.D. Department of Psychology Division Neuropsychology Binzmuehlestrasse 14, PO Box 25 8050 Zurich, Switzerland

0041 44 635 73 97 (phone office) 0041 76 445 86 84 (phone mobile) 0041 44 635 74 09 (fax office) BIN 4.D.04 (office room number)

j.haenggi@psychologie.uzh.ch (email) http://www.psychologie.uzh.ch/neuropsy/ (website) http://www.juergenhaenggi.ch (private website)

This e-mail (and any attachment/s) contains confidential and/or privileged information. If you are not the intended recipient (or have received this e-mail in error) please notify the sender immediately and destroy this e-mail. Any unauthorised copying, disclosure or distribution of the material in this e-mail is strictly forbidden.

---

---

Am 13.10.2016 um 15:56 schrieb Eelco van Duinkerken:

Hi all,

I am using FS with data from 2 different studies that were acquired on the same MRI-machine with the same T1 and FLAIR sequences. Unfortunately, the group sizes are not very balanced, with 31 controls, 16 obese and 32 diabetes patients.

Is the GLM for thickness used in FS very sensitive to this unequal group size?

Thanks for the help,

Eelco _______________________________________________ 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.

---

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.