• If you are citizen of an European Union member nation, you may not use this service unless you are at least 16 years old.

  • Stop wasting time looking for files and revisions. Connect your Gmail, DriveDropbox, and Slack accounts and in less than 2 minutes, Dokkio will automatically organize all your file attachments. Learn more and claim your free account.


Analysis 2: Multiple Regression

Page history last edited by PBworks 13 years, 6 months ago

Analysis 2: Multiple Regression

(1) To do this analysis, you'll need a Design Matrix file. You can download a Design Matrix file specifically for the IR1 (should work for IR2 or any other paradigm that follows the same block sequence) here; save it to the tutorial directory.


You might be wondering where the Design Matrix file came from. If you'd like to see an example of how to make a Design Matrix file for IR1, see Make Design Matrix for IR1.


(2) Open up a new MEDx session, and then open up a New Folder.


(3) Load the Mask image that you created in step 12 of Spatial Normalization (the one named SnMask.hdr, not Mask.hdr).


(4) Also load the 84 spatially normalized, spatially smoothed, temporally filtered, motion-corrected image volumes which you created in the Spatial Normalization step (the ones named Sn*.hdr). You should now be looking at a Group page showing the 84 spatially normalized scans. The name of the Group page should be New Group.


(5) Select Toolbox --> Functional --> Correlation Analysis....


(6) In the Correlation Analysis dialog box, click on the Multiple Regression tab.


(7) Set Time Series Group to New Group (which is the name of the Group page you want to do the analysis on), and Input Model File to the DesignMatrix.vst file you either downloaded or created in step 1. For the latter, the Input Model File entry field should have the full path name to the design matrix file, e.g.




You can use the Select... or browser buttons if you like.


(8) Under Generate Images, unselect Model Probability and Parameter Probability, since we rarely use these. Leave Model F and r^2 and Parameter Estimate, t and Z-score set.


(9) Leave Apply Temporal Autocorrelation Correction set. Set Hemodynamic Lag (secs) to 5, and TR (secs) to 3. (The use of the term lag here is a little misleading; it probably should have been spread instead, since it is the hemodynamic spread rather than lag that is the cause of the temporal autocorrelation, which in turn requires an adjustment of degrees of freedom.) Then click on the OK button. This will cause the Multiple Regression analysis to be performed.


(10) Examine the results. Click on the button labeled MR: Z for Custom - 1, df = 77, which is the Z map for the contrast "(RealWord and FalseFont) Minus Fix". Select Display --> Display Range..., and and determine the minimum and maximum values of the image as demonstrated in step 6 of Obtain data for tutorial via DICOM transfer. Select Image --> Image Properties..., and determine the image dimensions, the pixel encoding, and the bits per pixel.


(11) Mask this Z map in the manner demonstrated in steps 13-17 of Analysis 1: Simple t-test.


(12) Return to the Results page of the multiple regression analysis, and then repeat steps 10 and 11 for the Z map labeled MR: Z for Custom - 2, df = 77, which is the Z map for the contrast "RealWord Minus FalseFont".


(13) The buttons labeled MR: Parameter Est for Custom - 1 and MR: Parameter Est for Custom - 2 are the parameter estimates (regression coefficients) for the contrasts "(RealWord and FalseFont) Minus Fix" and "RealWord Minus FalseFont".


(14) Save the two Z maps into two AVW format files named Z.multregress.RealWordAndFalseFontVsFix.hdr and Z.multregress.RealWord-FalseFont.hdr.


(15) Exit MEDx.


The next step is Cluster Detection.


See also this Multiple Regression Tutorial.


If you really want to get into the nuts and bolts of setting up design matrices, see this Tutorial on the Construction of Design Matrices. I believe I am the author this tutorial -- I probably wrote it when I was at Sensor Systems the first time. At least, the writing style seem to be very similar to my own.


Return to tutorial main page

Comments (0)

You don't have permission to comment on this page.