External Email - Use Caution
Dear Freesurfer Experts, Hope this finds you well
I am following the LME tutorial running LME Univariate analysis. I was wondering if someone would kindly check my steps below? I wanted to make sure specially that my matrix (X) and and contrast (C) were designed accordingly to assess group*time interaction.
Study design: Two groups: 0 = control, 1 = intervention Time: 0 = baseline, 1 = 6-month follow-up Outcome of interest: Change in hippocampal volume Co-Variate: Estimated ICV
Here are the steps I took in Matlab:
# Loading into Matlab Qdec table with 7 columns in this exact order: fsid, fsid-base, time, group, Righ.Hipp, Left.Hipp, estimated_ICV: Qdec = fReadQdec('qdec-aseg.table.dat'); Qdec = rmQdecCol(Qdec,1); sID = Qdec(2:end,1); Qdec = rmQdecCol(Qdec,1); M = Qdec2num(Qdec); Y = M(:,3:4); M = M(:,[1:2 5]); # Here I loaded estimated ICV already into M (as the third column), as it will be treated as a covariate later.
# Sorting data according to time [M,Y,ni] = sortData(M,1,Y,sID);
# Building design matrix X from M, for a simple time by group interaction. Note that here, I have the slope column as 1, then I add time and group columns via: M(:,1:2), then a column for group*time via: M(:,1).*M(:,2), and finally a column for estimated ICV, as the final column in the matrix via: M(:,3). I wonder if this is correct? X = [ones(length(M),1) [M(:, 1:2) M(:,1).*M(:,2) M(:,3)]];
# Determining model parameters, for total hippocampal volume, Left.Hipp, and Right.Hipp. Here, since my sample is small (N = 15), I am using 1 random effect only. hipp_vol_stats_tot = lme_fit_FS(X,[1],Y(:,1)+Y(:,2),ni); hipp_vol_stats_left = lme_fit_FS(X,[1],Y(:,1),ni); hipp_vol_stats_right = lme_fit_FS(X,[1],Y(:,2),ni);
# Creating contrast (C) matrix for simple group*time interaction controlling for estimated ICV. Does this look correct, considering that the last 0 would control for estimated ICV? C = [ 0 0 0 1 0];
# Finally, F test for total, left and right hippocampal volumes F_C_tot = lme_F(hipp_vol_stats_tot,C); F_C_left = lme_F(hipp_vol_stats_left,C); F_C_right = lme_F(hipp_vol_stats_right,C);
Thanks so much for your help!
Nárlon Cássio
Nárlon Cássio Boa Sorte Silva, PhD CIHR/MSFHR Postdoctoral Research Fellow Aging, Mobility, and Cognitive Neuroscience Lab Djavad Mowafaghian Centre for Brain Health University of British Columbia Twitter: @BoaNarlon
External Email - Use Caution
Just following up on my email below to see if anyone is able to help?
Thanks! Nárlon Cássio
Nárlon Cássio Boa Sorte Silva, PhD CIHR/MSFHR Postdoctoral Research Fellow Aging, Mobility, and Cognitive Neuroscience Lab Djavad Mowafaghian Centre for Brain Health University of British Columbia Twitter: @BoaNarlon
On Jan 20, 2021, at 9:20 PM, Boa Sorte Silva, Narlon <narlon.silva@ubc.camailto:narlon.silva@ubc.ca> wrote:
[CAUTION: Non-UBC Email] External Email - Use Caution
Dear Freesurfer Experts, Hope this finds you well
I am following the LME tutorial running LME Univariate analysis. I was wondering if someone would kindly check my steps below? I wanted to make sure specially that my matrix (X) and and contrast (C) were designed accordingly to assess group*time interaction.
Study design: Two groups: 0 = control, 1 = intervention Time: 0 = baseline, 1 = 6-month follow-up Outcome of interest: Change in hippocampal volume Co-Variate: Estimated ICV
Here are the steps I took in Matlab:
# Loading into Matlab Qdec table with 7 columns in this exact order: fsid, fsid-base, time, group, Righ.Hipp, Left.Hipp, estimated_ICV: Qdec = fReadQdec('qdec-aseg.table.dat'); Qdec = rmQdecCol(Qdec,1); sID = Qdec(2:end,1); Qdec = rmQdecCol(Qdec,1); M = Qdec2num(Qdec); Y = M(:,3:4); M = M(:,[1:2 5]); # Here I loaded estimated ICV already into M (as the third column), as it will be treated as a covariate later.
# Sorting data according to time [M,Y,ni] = sortData(M,1,Y,sID);
# Building design matrix X from M, for a simple time by group interaction. Note that here, I have the slope column as 1, then I add time and group columns via: M(:,1:2), then a column for group*time via: M(:,1).*M(:,2), and finally a column for estimated ICV, as the final column in the matrix via: M(:,3). I wonder if this is correct? X = [ones(length(M),1) [M(:, 1:2) M(:,1).*M(:,2) M(:,3)]];
# Determining model parameters, for total hippocampal volume, Left.Hipp, and Right.Hipp. Here, since my sample is small (N = 15), I am using 1 random effect only. hipp_vol_stats_tot = lme_fit_FS(X,[1],Y(:,1)+Y(:,2),ni); hipp_vol_stats_left = lme_fit_FS(X,[1],Y(:,1),ni); hipp_vol_stats_right = lme_fit_FS(X,[1],Y(:,2),ni);
# Creating contrast (C) matrix for simple group*time interaction controlling for estimated ICV. Does this look correct, considering that the last 0 would control for estimated ICV? C = [ 0 0 0 1 0];
# Finally, F test for total, left and right hippocampal volumes F_C_tot = lme_F(hipp_vol_stats_tot,C); F_C_left = lme_F(hipp_vol_stats_left,C); F_C_right = lme_F(hipp_vol_stats_right,C);
Thanks so much for your help!
Nárlon Cássio
Nárlon Cássio Boa Sorte Silva, PhD CIHR/MSFHR Postdoctoral Research Fellow Aging, Mobility, and Cognitive Neuroscience Lab Djavad Mowafaghian Centre for Brain Health University of British Columbia Twitter: @BoaNarlon
_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edumailto:Freesurfer@nmr.mgh.harvard.edu https://secure-web.cisco.com/1qEruuMyPPYa8DnjG0x-Dj-Ko09YcKFDsc58L4q-PQhpCGv...
freesurfer@nmr.mgh.harvard.edu
-
Boa Sorte Silva, Narlon