Index

Name

mri_gdfglm - performs GLM analysis given group descriptor file (GDF) and dependent variable table (DVT)

Synopsis

mri_gdfglm --gdf <gdfile> --dvt <dvtfile> [options]

Arguments

Required Flagged Arguments

--gdf gdfile

group descriptor file

Path to the GDF. See http://surfer.nmr.mgh.harvard.edu/docs/fsgdf.txt for more info. This file will have a list of Classes and Variables. Hereafter, the Variables are referred to as Covariates.

--dvt dvtfile

dependent variable file

Path to the dependent variable table (DVT). This is a text file that contains the data table. The first column is the list the names of each row. The first row is the list of the names of each column. There needs to be a text place-holder at the first row/column. The rest of the table is filled with numbers. Each column should be a subject, and each row an observation for that subject. The arguments of --depvar must come from the row names. A DVT is produced by make-segvol-table.

Optional Flagged Arguments

--testoffset

test offset, not covariate slope

--keepid

print subjid in output.dat

--o

output base name

Base name of output files. There will be four output files created: (1) the summary file (basename.sum), (2) the data file (basename.dat), (3) the GDF (basename.gdf), and a matrix file (basename.mat). The summary file has a list of the parameters used in the analysis as well as the results, including parameter estimates, contrast effect size, and signficance. The data file contains a table of the final data with each subject on a different row. The first column is the subject number, the next nCV are the nCV covariates, the next column is the final dependent variable, and the final column is the best fit of the dependent variable. The GDF is the final GDF; this will be the same as the input GDF if no classes, covariates or subjects have been excluded. The matfile is a matrix in matlab4 format that contains the design matrix concatenated with the final dependent variable, the fit of final dependent variable, and the residual. In addition, each class has its own output dat file called outbase-classlabel.dat. The first column is the subject number, the next nCV are the nCV covariates, the next column is the final dependent variable, and the final column is the best fit of the dependent variable. This output is best for creating scatter plots.

--classes Class1 Class2

use subset of classes

Use only the subjects that belong to the specfied list of classes. The class names must come from the classes as specified in the GDF. If unspecfied, all classes are used.

--covar Covar1 Corvar2

use subset of covars

Use only the variables that belong to the specfied list. The names must come from the variables as specified in the GDF. If unspecfied, all variables are used.

--depvar DepVar1 <DepVar2 ...>

spec dependent variables

Select variables from the DVT. The a weighted average of variables will be computed as the final dependent variable. If unspecified, all variables will be used. See --wdepvar.

--wdepvar wdv1 wdv2

weight depvars (default is 1)

Set the weights of the dependent variables. The final dependent variable will be computed as a weighted average of the dependent variables. The number of weights must be equal to either the number of DepVars listed in --depvar or (if --depvar is unspecfied) the number of dependent variables in the DVT. If unspecfied, the weights are set to compute a simple average.

--wclass wc1 wc2

Class weights (default is 1)

Class weights for establishing a contrast. The number of weights must be equal to the number of classes (ie, the number listed in --classes or the number in the GDF). If unspecified, all weights are set to 1. This applies only to the contrast; if the weight of a class is set to 0, that class is still included in the parameter estimation. If positive and negative weights are used, they should sum to the same value.

--wcovar wcv1 wcvw

Covar slope weights

Covariate weights for establishing a contrast. The number of weights must be equal to the number of covariates (ie, the number listed in --covar or the number in the GDF). If unspecified, all weights are set to 1. This applies only to the contrast; if the weight of a covariate is set to 0, that covariate is still included in the parameter estimation. If positive and negative weights are used, they should sum to the same value.

Outputs

??

Description

This command performs glm analysis given group descriptor file (GDF) and dependent variable table (DVT).

Examples

Example 1

Consider the following Group Descriptor File:

------------------------------------------
GroupDescriptorFile 1
  Title AlbertGroup
  Class NormMale   
  Class AlzMale    
  Class NormFemale 
  CLASS AlzFemale  
  DefaultVariable Age
  Variables   Age MMSE
Input 003007 AlzMale 75 30 
...
------------------------------------------

Test whether the left hippocampal volume signficantly varies with MMSE for the Alzheimer group regressing out the effect of gender and age:

mri_gdfglm --gdf albert.gdf --dvt asegvol.dat --o lhipmmse 
  --wclass 0.5 -0.5 0.5 -0.5  --wcovar 0 1 
  --depvar Left-Hippocampus 

Example 2

Test whether the left-right difference in hippocampal volume signficantly varies across gender using only the normal subjects, and regressing out the effect of age but not MMSE.

mri_gdfglm --gdf albert.gdf --dvt asegvol.dat --o hip-lrdiff-mvf 
  --depvar Left-Hippocampus Right-Hippocampus --wdepvar  1 -1 
  --classes NormMale NormFemale   --wclass  1 -1  --covar Age 
  --testoffset 

Bugs

None

See Also

??

Links

Freesurfer GDF, GLM

Methods Description

N/A

References

N/A

Reporting Bugs

Report bugs to <analysis-bugs@nmr.mgh.harvard.edu>

Author/s

DougGreve

mri_gdfglm (last edited 2008-04-29 11:45:32 by localhost)