Dear Alfonso

Since we are part of a multi-center study, we have received a lot of resting state fMRI data (n>150) from different sites (n=9) that we plan to analyze. 
At the moment, we are having three covariates of no interest (age, gender, IQ) for our sample of patients and controls besides some covariates of interest (e.g., ADHD symptoms). We centered "age" and "IQ" values, and defined "gender" as "1=male", "2=female".


My question is: How do we define "site" as covariate of no interest in order to add "site" to our "2nd level covariates"?


Thanks a lot for your help!


Best regards
Julia

Hello,

may I follow up on the question of the proper specification of the covariates? Are any covariates automatically mean-centered in Conn before the 2nd-level analysis?

For example, the QA variables are often included as covariates of no interest as well as the age. However, on the forum it is often advised to mean-center the Age covariate, but the QA variables are usually not considered. Does it mean that they are mean-centered automatically? Or one should better demean them before usage like any other covariate, e.g. Age or IQ?

Thank you!

Best regards,
Ekaterina.

Hi Conners, dear Alfonso

I just found a post by Alfonso regarding my question.
It is from february 20th, 2017 (citation): "Regarding QA_GCOR that is an automatic Quality Assurance covariate computed during the Denoising step. It represents the average global correlation (GCOR) for each subject and each condition (i.e. the center of the voxel-to-voxel histograms that are displayed in the Denoising tab) which has been proposed/used in several studies as a control covariate when comparing groups of subjects across different sites, or when needing additional control for potential noise/artifactual effects on your connectivity measures."

Thanks a lot, Alfonso, this helped!

Kind regards
Julia

Hi Julia,

thank you very much for sharing!

But anyway would be great to know how one would proper specify in Conn the categorical covariates with more than two categories and whether one needs to demean the QA variables by hand or not.

Kind regards,
Ekaterina.


Originally posted by Julia Werhahn:

Dear Conn users, dear Alfonso,

I am currently analysing rs-fMRI from 3 different sites (scan from different manufacturers but all with exactly the same scan parameters). I have some troubles to correct for the site effect by using conn.
I have read the post suggesting that GCOR could be used when comparing subjects across different sites and I have some questions :

1) Is there any other way to correct the site effect by using categorical 2nd level covariates ?

2) If I use GCOR, should I just add this covariate in the denoising step ? Is there any reference that can be cited to justify this method ?

Many thanks in advance for your help,

Best wishes,

Ali

Hi, I am also looking for answer on how to properly control for multiple sites in the contrast manager.

Currently I have 5 sites and have dummy coded them as 5 separate variables.


If I want to do an resting state ancova controlling for age, sex, sites would it look like this?

Group Covariate Age Sex Site1 Site2 Site3 Site4 Site5

 [  0        1         0     0      0     0       0       0      0   ]

Originally posted by Ali Amad:

Dear Ali, conn users and experts,
I wonder if you have found an answer to your question regarding how to correct the multi site (scanner) effect, probably through defining a categorical second level covariate?
Thanks in advance!

Originally posted by Zahra Mor:

Hey People!

I have the same problem!!!!

Anybody found a solution?

best wishes
Till

Hi Till,

Exactly, simply define a set of site-specific covariates (e.g. SITE_1, SITE_2, etc.) and include these as covariates-of-no-interest in your second-level analysis (e.g. image attached). 

For example, if you have three sites and have already imported a 2nd-level SITE covariate with values 1 to 3 indicating the site of each subject, you could: 

1) discretize that variable: in the Setup.Covariates.2nd-level tab select your covariate SITE, and then click on 'Covariate tools. Discretize selected covariate'. That will create three new covariates named SITE_1, SITE_2 and SITE_3

2) (optionally / rarely-necessary) center those new site-specific covariates to your desired control-level (e.g. average across all subjects):  in the same tab select jointly SITE_1, SITE_2 and SITE_3, then click on 'Covariate tools. Orthogonalize selected covariates', and select the variable 'AllSubjects' as your only orthogonal factor

3) add those site-covariates as controls in your second-level analysis. For example, in the Results (2nd-level) tab, after defining your desired analysis (e.g. a two-sample t-test comparing PLACEBO and TREATMENT subjects), simply select the option that reads 'add/remove SITE_1 as control covariate' (and repeat for SITE_2 and SITE_3). You should be seeing something like in the example attached, which will estimate the differences between groups while controlling for site effects. 

Hope this helps
Alfonso
 
Originally posted by Till Langhammer:

Hi Till and Alfonso,

I was wondering whether all the sites in your data had same fMRI acquisition protocol and scanner models (such as GE / Philips etc)?

I am also looking into combining the data acquired from different scanner models (Philips + Siemens) and of course the protocols are not exactly the same. Can the multi-site 2nd-level covariate correction method be applied to the data with different scanner brands and MRI protocol? The data I have is both structural and functional scans from dementia cohorts and the aim is to combine the matching subjects from different sites into one fMRI analysis pipeline and investigate the different hypothesis?

Thank you!,

Regards,
Dilip Kumar

Hi Alfonso,


I have a peculiar case trying to control for site for two groups. All subjects in group A came from site A and all subjects in group B came from site B. i want to do a -1 1 contrast to look at differences between groups.

When I do option 3) mentioned in your post do do this between group contrast, I get a "Warning" (see screenshot: possible incorrect model, non-estimable contrast. suggestion simplify second-level model). What can I do to go around this or if there is another way I can control for site difference.

Thank you so much,
Nancy

Originally posted by Alfonso Nieto-Castanon:

Hi Nancy,

Yes, unfortunately the warning correctly indicates that you cannot possibly control for site-effects if those are exactly co-linear with your group-effects (e.g. as in your case if all group A subjects came from site A and all group B subjects came from site B). When looking at group-effects while controlling for site-effects the analyses will effectively perform this control by estimating the differences in connectivity between the two groups when compared separately within each site (e.g. groupA - groupB differences restricted to only subjects from site1, groupA - groupB differences from subjects in site2, etc.), so the warning is just indicating that unfortunately those comparisons simply cannot be made in your case. 

Best
Alfonso
Originally posted by Nancy Mugisha:

Hi Alfonso,

Is this different than creating a variable called "site" with 3 levels (1,2,3), and including that as a covariate of no interesT?

Thank you,
Alex

Hi Alex,

While that approach is fine for two sites (or when controlling for the effect of a categorical factor with two levels), for three or more sites the appropriate control will involve defining one covariate per site, each with 0/1 values indicating which subjects come from each individual site. Using instead as a control a variable with values 1,2,3 will only control for sorted effects across those three sites, leaving other potential inter-site differences uncontrolled.

Best
Alfonso
Originally posted by Alexandra Muratore:

Hi Alfonso,

I have a question about this issue of study sites as covariates in CONN (see your comment below). What you described sounds to me like a full dummy coding with n-covariates for n-scanning sites and each participant scanned at these sites is coded with 1 and the others with 0. Is it possible in CONN to dummy code the scanning sites with a reference group (see figure below) with n-1 covariates?

source: https://de.mathworks.com/help/stats/dumm...

Or is it more common to code the scanning sites as full dummy variables rather than dummy variables with a reference group?

Thank you so much for your help!

 

Best,

Max

 

Originally posted by Alfonso Nieto-Castanon:

Hi Till, Exactly, simply define a set of site-specific covariates (e.g. SITE_1, SITE_2, etc.) and include these as covariates-of-no-interest in your second-level analysis (e.g. image attached).  For example, if you have three sites and have already imported a 2nd-level SITE covariate with values 1 to 3 indicating the site of each subject, you could:  1) discretize that variable: in the Setup.Covariates.2nd-level tab select your covariate SITE, and then click on 'Covariate tools. Discretize selected covariate'. That will create three new covariates named SITE_1, SITE_2 and SITE_3 2) (optionally / rarely-necessary) center those new site-specific covariates to your desired control-level (e.g. average across all subjects):  in the same tab select jointly SITE_1, SITE_2 and SITE_3, then click on 'Covariate tools. Orthogonalize selected covariates', and select the variable 'AllSubjects' as your only orthogonal factor 3) add those site-covariates as controls in your second-level analysis. For example, in the Results (2nd-level) tab, after defining your desired analysis (e.g. a two-sample t-test comparing PLACEBO and TREATMENT subjects), simply select the option that reads 'add/remove SITE_1 as control covariate' (and repeat for SITE_2 and SITE_3). You should be seeing something like in the example attached, which will estimate the differences between groups while controlling for site effects.  Hope this helps Alfonso   Originally posted by Till Langhammer:

Originally posted by Zahra Mor:

Dear Ali, conn users and experts, I wonder if you have found an answer to your question regarding how to correct the multi site (scanner) effect, probably through defining a categorical second level covariate? Thanks in advance!

Hey People! I have the same problem!!!! Anybody found a solution? best wishes Till

 

Hi Max,

Yes, you are exactly right, both approaches are perfectly fine (and using a reference site can make the specification of your contrast vector/matrix simpler in may cases).

Just to elaborate a bit, for example assuming that want to estimate patients vs control differences using data from three different sites it is equivalent to test a model of the form:

  Subject effects: Patients, Controls, Site1, Site2, Site3

  Between-subjects contrast: [1, -1, 0, 0, 0]

and a second model of the form:

  Subject effects: Patients, Controls, Site1, Site2

  Between-subjects contrast: [1, -1, 0, 0]

(where "Site1" is a variable with 1's for subjects in site 1 and 0's for all other subjects). The two models are equivalent because both design matrices span the same column space (only the first model design matrix contains a "redundant" column as the sum Patients+Controls is equal to the sum Site1+Site2+Site3), and both contrasts define the same dimension within that column space.

Even though both forms are exactly equivalent (and they will produce the exact same results/statistics),  what makes the second form desirable in many cases is that it simplifies the interpretation and specification of some common post-hoc contrasts. For example, if you knew that you wanted to use site3 as reference and would like to evaluate the adjusted means in Patients and Controls at the level of this reference site, in the first analysis you would do so by specifying a contrast of the form:

  Between-subjects contrast: [1, 0, 0, 0, 1] (patients adjusted means at level of site3)

  Between-subjects contrast: [0, 1, 0, 0, 1] (controls adjusted means at level of site3)

while in the second analysis you would do so by specifying a contrast of the form:

  Between-subjects contrast: [1, 0, 0, 0] (patients adjusted means at level of site3)

  Between-subjects contrast: [0, 1, 0, 0] (controls adjusted means at level of site3)

with the latter being simpler as the model regressor coefficients directly inform you of the adjusted means within each group at the desired control level. 

Hope this helps

Alfonso

 

Originally posted by max345:

Hi Alfonso,

I have a question about this issue of study sites as covariates in CONN (see your comment below). What you described sounds to me like a full dummy coding with n-covariates for n-scanning sites and each participant scanned at these sites is coded with 1 and the others with 0. Is it possible in CONN to dummy code the scanning sites with a reference group (see figure below) with n-1 covariates?

source: https://de.mathworks.com/help/stats/dumm...

Or is it more common to code the scanning sites as full dummy variables rather than dummy variables with a reference group?

Thank you so much for your help!

 

Best,

Max

 

Originally posted by Alfonso Nieto-Castanon:

Hi Till, Exactly, simply define a set of site-specific covariates (e.g. SITE_1, SITE_2, etc.) and include these as covariates-of-no-interest in your second-level analysis (e.g. image attached).  For example, if you have three sites and have already imported a 2nd-level SITE covariate with values 1 to 3 indicating the site of each subject, you could:  1) discretize that variable: in the Setup.Covariates.2nd-level tab select your covariate SITE, and then click on 'Covariate tools. Discretize selected covariate'. That will create three new covariates named SITE_1, SITE_2 and SITE_3 2) (optionally / rarely-necessary) center those new site-specific covariates to your desired control-level (e.g. average across all subjects):  in the same tab select jointly SITE_1, SITE_2 and SITE_3, then click on 'Covariate tools. Orthogonalize selected covariates', and select the variable 'AllSubjects' as your only orthogonal factor 3) add those site-covariates as controls in your second-level analysis. For example, in the Results (2nd-level) tab, after defining your desired analysis (e.g. a two-sample t-test comparing PLACEBO and TREATMENT subjects), simply select the option that reads 'add/remove SITE_1 as control covariate' (and repeat for SITE_2 and SITE_3). You should be seeing something like in the example attached, which will estimate the differences between groups while controlling for site effects.  Hope this helps Alfonso   Originally posted by Till Langhammer:

Originally posted by Zahra Mor:

Dear Ali, conn users and experts, I wonder if you have found an answer to your question regarding how to correct the multi site (scanner) effect, probably through defining a categorical second level covariate? Thanks in advance!

Hey People! I have the same problem!!!! Anybody found a solution? best wishes Till

 

 

Dear Alfonso,

I have specified my three sites as three covariates: site 1, 2 and 3 with values 0 and 1. I want to look at the effect of AllSubjects. If I include all three of them [1 0 0 0], I get the warning: 'possibly incorrect model: non-estimable contrasts (suggestion: simplify second-level model). As I understood from your answers in this thread, it should not matter when I leave site 1, 2 or 3 out. However, I do get different results when I leave one of them out, and it also depends on which one I leave out.. Can you help me understand what is going wrong here?

Best regards,
Nikki