Differences

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

--- processing_stream_-_scripted [2010/06/09 20:45] – jochen
+++ processing_stream_-_scripted [2010/06/09 21:09] – added link/reference to MDM::ComputeGLM jochen
@@ Line 168: / Line 168: @@
     end
 end</code>
+===== Semi-automatic logfile parsing and PRT creation =====
+The onsets have supposedly already been transformed into a tabular structure (e.g. in Microsoft Excel), so that they look like this:
+<file>SUBJ_ID	INST_on	CUE1_on	CUE2_on	INST_of	CUE1_of	CUE2_of
+	3	nan	nan	5003	nan	nan
+	nan	3003	nan	nan	18003	nan
+	18000	nan	nan	23000	nan	nan
+	nan	nan	21000	nan	nan	26000
+	74057	nan	nan	79057	nan	nan
+	nan	77057	nan	nan	92057	nan
+	92068	nan	nan	97068	nan	nan
+	nan	nan	95067	nan	nan	100067
+	148112	nan	nan	153112	nan	nan
+	nan	151112	nan	nan	166112	nan
+	166122	nan	nan	171122	nan	nan
+	nan	nan	169122	nan	nan	174122
+	1	nan	nan	5001	nan	nan
+	nan	3001	nan	nan	18001	nan
+	17999	nan	nan	22999	nan	nan
+	nan	nan	21011	nan	nan	26011
+	74056	nan	nan	79056	nan	nan
+	nan	77056	nan	nan	92056	nan
+	92066	nan	nan	97066	nan	nan
+	nan	nan	95065	nan	nan	100065
+	148110	nan	nan	153110	nan	nan
+	nan	151110	nan	nan	166110	nan
+	166120	nan	nan	171120	nan	nan
+	nan	nan	169120	nan	nan	174120	</file>
+In this table, the first column codes for the subject ID (which is used to detect when next subject occurs, in this sample there is only one subject!). The next three columns code the onset and the final three columns the offset of the three conditions used in this experiment (INST = Instruction, CUE1 = cue type 1, CUE2 = cue type 2).
+The following script will then create a set of PRT files out of this table (which can be copied into Matlab):
+<code matlab SCANUnit_CPU_prtfromtable.m>% make sure the onset table is available in variable ot !!
+% configuration
+conds = {'INST', 'CUE1', 'CUE2'};
+condcol = [255, 0, 0; 0, 255, 0; 0, 0, 255];
+% find unique subject IDs
+subjid = unique(ot(:, 1));
+% for each subject ID
+for sc = 1:numel(subjid)
+    % get this subject's sub-table
+    sot = ot(ot(:, 1) == subjid(sc), :);
+    % set NaNs to zero (for now)
+    sot(isnan(sot)) = 0;
+    % get onsets and offsets of any trial
+    onsets = sum(sot(:, 2:2+numel(conds)-1), 2);
+    % get table with NaNs (again) for condition detection
+    sot = ot(ot(:, 1) == subjid(sc), :);
+    % find run-separators
+    runseps = 1 + [0; find(diff(onsets) < 0)];
+    runseps(end+1) = numel(onsets) + 1;
+    % for each run
+    for rc = 1:numel(runseps)-1
+        % get part of table we need
+        rot = sot(runseps(rc):runseps(rc+1)-1, 2:end);
+        % create new PRT
+        prt = xff('new:prt');
+        % for each condition
+        for cc = 1:numel(conds)
+            % add to PRT
+            prt.AddCond(conds{cc}, rot(~isnan(rot(:, cc)), [cc, cc + numel(conds)]), condcol(cc, :));
+        end
+        % save PRT
+        prt.SaveAs(sprintf('%d_%d.prt', subjid(sc), rc));
+        % clear object
+        prt.ClearObject;
+    end
+end</code>
+**Note: the PRT files will be written into the present working directory (unless the SaveAs line is appropriately altered), and they need to be moved or copied into their respective subject's folder!**
+===== Computing an RFX-GLM with VTCs and PRTs =====
+Once all files are in place, the Protocol and VTC files can be used to compute an RFX-GLM. Please follow the instructions given in the example of the [[mdm.ComputeGLM]] method reference page.