User Tools

Site Tools


glm.singlestudy_tmap

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

glm.singlestudy_tmap [2010/08/24 15:39] (current)
jochen created
Line 1: Line 1:
 +====== GLM::​SingleStudy_tMap ======
  
 +===== Motivation =====
 +After running the regression for a single study (run, i.e. one FMR, VTC, or MTC object), creating t-contrast maps for just this study can be beneficial in determining whether or not a certain run contains too much noise (or specific artefacts) for it to be included in second-level statistics. Also, it is sometimes desired to use the actually standard-error-normalized effect-statistic (t instead of regression beta) for further computations (e.g. to incorporate a measure of within-subject-data noise on the second level).
 +
 +===== Requirements =====
 +You need to have a GLM file/object available that represents the regression outcome of a single run. **Please note that, at this time, the support for FMR (MAP creation) has not been implemented;​ but as the MAP format lacks a lot of the properties (and thus flexibility) of the VMP format, it is also not suggested other than for very specific application,​ such as MVPA.**
 +
 +===== Method reference ('​glm.Help('​SingleStudy_tMap'​)'​) =====
 +<​file> ​ GLM::​SingleStudy_tMap ​ - calculate a t contrast map
 +
 +  FORMAT: ​      map = glm.SingleStudy_tMap([c,​ mapopts])
 +
 +  Input fields:
 +
 +        c           NxC contrast vector (default: full model and main eff)
 +        mapopts ​    ​structure with optional fields
 +         ​.interp ​   mesh-based interpolation (default: true)
 +         ​.srf ​      ​surface file, required for interpolation
 +
 +  Output fields:
 +
 +        map         ​MAP/​VMP/​SMP object with C maps</​file>​
 +
 +===== Reference notes =====
 +**The ''​.interp''​ option** was intended to cover those cases where in (fairly "​old"​ versions of BrainVoyager QX, 1.7.x), vertex nodes would sometimes lack an appropriate target when a Sphere-to-Sphere-Mapping (SSM) object had been specified. As this bug has been fixed (and a workaround is still available on the SMP side), this option **will be removed in future versions**.
 +
 +===== Usage example =====
 +Say you have a study with 5 regressors of interest and 1 confound (mean study level, automatically added by BrainVoyager QX/​NeuroElf),​ whereas the conditions of interest are:
 +
 +  * instruction
 +  * motion in left visual field
 +  * motion in right visual field
 +  * motion in both visual fields (at the same time)
 +  * static (trials without any motion)
 +
 +then the syntax to create a contrast over all conditions sharing motion in any visual field would be coded as:
 +
 +<code matlab glm_singlestudy_tmap_ex1.m>​% load a GLM (only needed if not yet loaded!)
 +glm = xff('​*.glm',​ '​Please select the single-study GLM...'​);​
 +
 +% create the contrast
 +contrast = glm.SingleStudy_tMap([0;​1;​1;​1;​0]);​
 +
 +% name the contrast
 +contrast.Map.Name = '​Motion in any visual field';​
 +
 +% save the contrast
 +contrast.SaveAs;​
 +
 +% clean up
 +contrast.ClearObject;​
 +glm.ClearObject;</​code>​
 +
 +===== Usage notes =====
 +Please note the following details about this method and the example:
 +
 +  * at this time, the name of the contrast(s) cannot be specified but has to be set after the call
 +  * multiple contrasts can be given whereas the number of contrasts is the number of columns in the weights argument
 +  * each contrast should either have its weights be all positive (or all negative, weighted contrast over baseline, or reversed) or sum up to 0 (weighted differential contrast between conditions);​ in other words, contrasts with weights of different sign where the weights do not sum up to 0 are invalid for direct hypothesis testing!
glm.singlestudy_tmap.txt ยท Last modified: 2010/08/24 15:39 by jochen