<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
  <head>
    <meta content="text/html; charset=ISO-8859-1"
      http-equiv="Content-Type">
  </head>
  <body bgcolor="#ffffff" text="#000000">
    Thank you Jan-Mathijs,<br>
    I actually did not think about that...<br>
    However, I did this just to remove the central blobs, independently
    of the effect it may cause in the different conditions. <br>
    <br>
    My aim is to use montecarlo non-parametric statistics afterwards to
    localize the sources. Don't you think there's gonna be a huge
    difference between nai-normalized and non-normalized datasets?<br>
    <br>
    <br>
    Thanks again!<br>
    <br>
    <br>
    <br>
    El 07/04/2011 14:23, jan-mathijs schoffelen escribió:
    <blockquote
      cite="mid:AA5DEE76-200E-45D0-BECF-71AE60A9C079@donders.ru.nl"
      type="cite">
      <div>Dear Marc,</div>
      <div><br>
      </div>
      <div>It seems from your question that you are interested in single
        trial reconstructed time series. If your purpose is to do a
        statistical comparison across a set of conditions I would not do
        a noise normalization at all. You mention that you used a
        spatial filter common to the three conditions. As a consequence,
        the estimate of the noise will be independent of the condition,
        so any normalization you would apply just leads to a scaling of
        the data, and will not change the outcome of your statistical
        test. </div>
      <div>Yet, I understand that it sometimes makes sense to do a
        normalization in order to be able to make a sensible
        visualization of the data (i.e. removing the big central blob).
        Also, in this case I would compute a descriptive statistic
        across the conditions, rather than normalizing with an estimate
        of the projected noise (which in FieldTrip is rather rudimentary
        because it assumes the noise to be spatially white). You could
        for example compute  an F-value from a one-factor, three level
        anova, or do a pairwise comparison of conditions using a t-test.</div>
      <div>Another alternative (which does not affect your statistical
        test either), is to work with norm-normalized leadfields. This
        will take away the blob in the centre of the volume, and
        facilitates visualization.</div>
      <div><br>
      </div>
      <div>Best wishes,</div>
      <div><br>
      </div>
      <div>Ja-Mathijs</div>
      <div><br>
      </div>
      <br>
      <div>
        <div>On Apr 5, 2011, at 6:54 PM, Marc Recasens wrote:</div>
        <br class="Apple-interchange-newline">
        <blockquote type="cite"><font class="Apple-style-span"
            face="arial, helvetica, sans-serif">Dear all,</font>
          <div><font class="Apple-style-span" face="arial, helvetica,
              sans-serif"><br>
            </font></div>
          <div><font class="Apple-style-span" face="arial, helvetica,
              sans-serif">As far as I know it is possible to reconstruct
              the time-course of the sources (obtained using
              sourceanalysis) by projecting/multiplying the
              filter-weights on the data.</font></div>
          <div><span class="Apple-style-span" style="white-space: pre;"><font
                class="Apple-style-span" face="arial, helvetica,
                sans-serif">In my case, the specific code I am using is:</font></span></div>
          <div><span class="Apple-style-span" style="white-space: pre;">
              <p class="p1"> <font class="Apple-style-span" size="2"
                  face="arial, helvetica, sans-serif"> <span class="s1">for</span>
                  i=1:nvoxels(inside)</font></p>
              <p class="p1"><font class="Apple-style-span" size="2"
                  face="arial, helvetica, sans-serif">        <span
                    class="s1">for</span> tr=1:ntrials</font></p>
              <p class="p1"><font class="Apple-style-span" size="2"
                  face="arial, helvetica, sans-serif">           
ts(tr,:)=(source.avg.ori{source.inside(i)}(1,:)*source.avg.filter{source.inside(i)}*data2{condit}.trial{tr});</font></p>
              <p class="p1"><font class="Apple-style-span" size="2"
                  face="arial, helvetica, sans-serif">        <span
                    class="s1">end</span></font></p>
              <p class="p1"><font class="Apple-style-span" size="2"
                  face="arial, helvetica, sans-serif">       
                  datvx(i,:,:)= single_trial_time_course</font></p>
              <p class="p1"><span class="Apple-style-span"
                  style="font-family: arial,helvetica,sans-serif;">end</span></p>
              <font class="Apple-style-span" face="arial, helvetica,
                sans-serif">I used a common filter with 3 different
                conditions, thus I think this is the only way I have to
                reconstruct the activity for one of the conditions.</font></span></div>
          <div><font class="Apple-style-span" face="arial, helvetica,
              sans-serif"><span class="Apple-style-span"
                style="white-space: pre;">My Question is about how to
                apply the NAI normalization here. Can I just divide the
                output of the source reconstruction by the projected
                noise of the filter? I am tempted to do something like
                that</span></font></div>
          <div><font class="Apple-style-span" face="arial, helvetica,
              sans-serif"><span class="Apple-style-span"
                style="white-space: pre;"><br>
              </span></font></div>
          <div><font class="Apple-style-span" face="arial, helvetica,
              sans-serif"><span class="Apple-style-span"
                style="white-space: pre;"></span></font><span
              class="Apple-style-span" style="font-family:
              arial,helvetica,sans-serif; white-space: pre;">datvx(i,:,:)=
              single_trial_time_course ./
              repmat(noise(inside),[size(ts,1), size(ts,2)]</span></div>
          <meta charset="utf-8">
          <div><span class="Apple-style-span" style="white-space: pre;"><font
                class="Apple-style-span" face="arial, helvetica,
                sans-serif"><br>
              </font></span></div>
          <div><span class="Apple-style-span" style="white-space: pre;"><font
                class="Apple-style-span" face="arial, helvetica,
                sans-serif">That is, I divide the power by the noise
                estimate (for each voxel) in every trial and time-point.
                Could anyone tell me whether this is a correct way to
                procedure? Any other workaround?</font></span></div>
          <div><span class="Apple-style-span" style="white-space: pre;"><font
                class="Apple-style-span" face="arial, helvetica,
                sans-serif"><br>
              </font></span></div>
          <div><span class="Apple-style-span" style="white-space: pre;"><font
                class="Apple-style-span" face="arial, helvetica,
                sans-serif"><br>
              </font></span></div>
          <div><span class="Apple-style-span" style="white-space: pre;"><font
                class="Apple-style-span" face="arial, helvetica,
                sans-serif">Thanks in advance!</font></span></div>
          <div><br>
          </div>
          <div>-- </div>
          <div> Marc Recasens<br>
            Tel.: +34 639 24 15 98<br>
            <br>
          </div>
          _______________________________________________<br>
          fieldtrip mailing list<br>
          <a moz-do-not-send="true"
            href="mailto:fieldtrip@donders.ru.nl">fieldtrip@donders.ru.nl</a><br>
          <a class="moz-txt-link-freetext" href="http://mailman.science.ru.nl/mailman/listinfo/fieldtrip">http://mailman.science.ru.nl/mailman/listinfo/fieldtrip</a></blockquote>
      </div>
      <br>
      <div apple-content-edited="true"> <span class="Apple-style-span"
          style="border-collapse: separate; color: rgb(0, 0, 0);
          font-family: Helvetica; font-style: normal; font-variant:
          normal; font-weight: normal; letter-spacing: normal;
          line-height: normal; orphans: 2; text-indent: 0px;
          text-transform: none; white-space: normal; widows: 2;
          word-spacing: 0px; font-size: medium;">
          <div style="word-wrap: break-word;"><span
              class="Apple-style-span" style="border-collapse: separate;
              color: rgb(0, 0, 0); font-family: Helvetica; font-size:
              medium; font-style: normal; font-variant: normal;
              font-weight: normal; letter-spacing: normal; line-height:
              normal; orphans: 2; text-indent: 0px; text-transform:
              none; white-space: normal; widows: 2; word-spacing: 0px;">
              <div style="word-wrap: break-word;">
                <div>Dr. J.M. (Jan-Mathijs) Schoffelen </div>
                <div>Donders Institute for Brain, Cognition and
                  Behaviour, <br>
                  Centre for Cognitive Neuroimaging,<br>
                  Radboud University Nijmegen, The Netherlands</div>
                <div><a moz-do-not-send="true"
                    href="mailto:J.Schoffelen@donders.ru.nl">J.Schoffelen@donders.ru.nl</a></div>
                <div>Telephone: 0031-24-3614793</div>
              </div>
            </span></div>
        </span> </div>
      <br>
      <pre wrap="">
<fieldset class="mimeAttachmentHeader"></fieldset>
_______________________________________________
fieldtrip mailing list
<a class="moz-txt-link-abbreviated" href="mailto:fieldtrip@donders.ru.nl">fieldtrip@donders.ru.nl</a>
<a class="moz-txt-link-freetext" href="http://mailman.science.ru.nl/mailman/listinfo/fieldtrip">http://mailman.science.ru.nl/mailman/listinfo/fieldtrip</a></pre>
    </blockquote>
    <br>
    <br>
    <pre class="moz-signature" cols="72">-- 
Marc Recasens
Tel.: +34 639 24 15 98</pre>
  </body>
</html>