 |
 |
. 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 -1 or alternatively, use some Matlab functions: [eye(9) 0-eye(9)] The trick is that if you get a significant difference, you don't know which time bin caused it, or which direction the difference was in (A>B or B>A), so you would need to investigate further to find out (e.g., plotting effects for each region). 2) A simpler approach is to simply look at the total "area under the curve" for each condition in a t-test. This effectively discards the timing information: [1 1 1 1 1 1 1 1 1 -1 -1 -1 -1 -1 -1 -1 -1 -1] for example would give you the sum of condition A compared to the sum of condition B (i.e., A > B). Alternatively you could write this [ones(1,9) ones(1,9)*-1]. Because this is a t-test it can give you information about directionality. However, you lose the temporal detail, which may be one of the reasons to choose an FIR model in the first place. Ultimately it will depend on what you are most interested in. There is some discussion of FIR models in Chapter 30 of the SPM manual (Face Group fMRI Data), which may also be of interest. Hope this helps, Jonathan -- Dr. Jonathan Peelle Department of Neurology University of Pennsylvania 3 West Gates 3400 Spruce Street Philadelphia, PA 19104 USA http://jonathanpeelle.net/ --0016363b959676d552049c59ca34 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable Hi Claire<div><div class="gmail_quote"><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">There are two conditions A and B I would like to compare.<br><br>In fMRI model specification, I chose to use the Finite Impulse Response such that my design has 9 three-second-bins for each condition.<br> <br> I would like to compare the two conditions in each time bin...is it possible to do this at once? Which of the t- or f-test should I be using and what should I put for contrasts? <br></blockquote><div><br></div><div>Two ways to compare A and B in your design are:</div> <div><br></div><div>1) Use an F test, where each row of the F test compares A vs. B at one time point. For example:</div><div><br></div><div>1 0 0 0 0 0 0 0 0 -1 0 0 0 0 0 0 0 0</div><div>0 1 0 0 0 0 0 0 0 0 -1 0 0 0 0 0 0 0</div> <div>0 0 1 0 0 0 0 0 0 0 0 -1 0 0 0 0 0 0</div><div>.</div><div>.</div><div>.</div><div>0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 -1</div><div><br></div><div>or alternatively, use some Matlab functions:</div><div><br></div><div>[eye(9) 0-eye(9)]</div> <div><br></div><div>The trick is that if you get a significant difference, you don't know which time bin caused it, or which direction the difference was in (A>B or B>A), so you would need to investigate further to find out (e.g., plotting effects for each region).</div> <div><br></div><div>2) A simpler approach is to simply look at the total "area under the curve" for each condition in a t-test. This effectively discards the timing information:</div><div><br></div><div>[1 1 1 1 1 1 1 1 1 -1 -1 -1 -1 -1 -1 -1 -1 -1]</div> <div><br></div><div>for example would give you the sum of condition A compared to the sum of condition B (i.e., A > B). Alternatively you could write this [ones(1,9) ones(1,9)*-1]. Because this is a t-test it can give you information about directionality. However, you lose the temporal detail, which may be one of the reasons to choose an FIR model in the first place.</div> <div><br></div><div>Ultimately it will depend on what you are most interested in. There is some discussion of FIR models in Chapter 30 of the SPM manual (Face Group fMRI Data), which may also be of interest.</div><div><br> </div><div>Hope this helps,</div><div>Jonathan</div><div><br></div><div><br>-- <br>Dr. Jonathan Peelle<br>Department of Neurology<br>University of Pennsylvania<br>3 West Gates<br>3400 Spruce Street<br>Philadelphia, PA 19104<br> USA<br><a href="http://jonathanpeelle.net/">http://jonathanpeelle.net/</a><br><br></div></div></div> --0016363b959676d552049c59ca34--