<html>
<head>
<style><!--
.hmmessage P
{
margin:0px;
padding:0px
}
body.hmmessage
{
font-size: 12pt;
font-family:Calibri
}
--></style></head>
<body class='hmmessage'><div dir='ltr'>Dear Eric Maris,<div><br></div><div>On the website I found a solution for calculating the interaction effect with permutation testing!</div><div>For my three way interaction I was thinking of the following solution and would like to know if this is statistically sound to do (perhaps also interesting for other people with the same design):</div><div><br></div><div>First I calculated the differences as proposed by you for an interaction effect, for <u>each intensity separately</u>:</div><div>Thus: </div><div><br></div><div><table border="0" cellpadding="0" cellspacing="0" width="226" style="border-collapse:
collapse;width:170pt">
<colgroup><col width="98" style="mso-width-source:userset;mso-width-alt:3584;width:74pt">
<col width="64" span="2" style="width:48pt">
</colgroup><tbody><tr height="20" style="height:15.0pt">
<td height="20" class="xl65" width="98" style="height:15.0pt;width:74pt"> </td>
<td class="xl67" width="64" style="border-left:none;width:48pt"><b>T0</b></td>
<td class="xl67" width="64" style="border-left:none;width:48pt"><b>T1</b></td>
</tr>
<tr height="20" style="height:15.0pt">
<td height="20" class="xl65" style="height:15.0pt;border-top:none"><b>Control arm</b></td>
<td class="xl66" style="border-top:none;border-left:none">A</td>
<td class="xl66" style="border-top:none;border-left:none">B</td>
</tr>
<tr height="20" style="height:15.0pt">
<td height="20" class="xl65" style="height:15.0pt;border-top:none"><b>Intervention arm</b></td>
<td class="xl66" style="border-top:none;border-left:none">C</td>
<td class="xl66" style="border-top:none;border-left:none">D</td>
</tr></tbody></table></div><div><br></div><div><br></div><div>B-A (with new variable <i>difference_B-A</i>) and C-D <span style="font-size: 12pt;">(with new variable</span><span style="font-size: 12pt;"> </span><i style="font-size: 12pt;">difference_C-D</i><span style="font-size: 12pt;">). I</span><span style="font-size: 12pt;">nstead of testing the interaction for these two new variables, </span></div><div><span style="font-size: 12pt;">I calculate the difference between </span><span style="font-size: 12pt;">variable</span><span style="font-size: 12pt;"> </span><i style="font-size: 12pt;">difference_C-D and </i><span style="font-size: 12pt;">variable</span><span style="font-size: 12pt;"> </span><i style="font-size: 12pt;">difference_B-A </i><font size="3">and use an F-statistic to express the differences across the 6 intensities and </font>subsequently<font size="3"> perform the permutation test. </font></div><div><br></div><div>Does this sound valid to you?</div><div>I'm really stuck, so hope that you will answer my question!</div><div><br></div><div>Best wishes,</div><div>Emanuel</div><div><br></div><div><br><div><hr id="stopSpelling">From: emanuelvandenbroeke@hotmail.com<br>To: fieldtrip@science.ru.nl<br>Subject: Cluster-based permutation test<br>Date: Thu, 19 Jun 2014 14:11:31 +0200<br><br>
<style><!--
.ExternalClass .ecxhmmessage P {
padding:0px;
}
.ExternalClass body.ecxhmmessage {
font-size:12pt;
font-family:Calibri;
}
--></style>
<div dir="ltr">Dear Eric Maris or other fieldtrippers, <div><br></div><div>In my experiment (a 2x2x6 design) I want to test whether ERP waveforms are significant different between conditions. There is no <i>a priori</i> assumption about the effects.</div><div>In this experiment I apply a somatosensory stimulus on both arms before and after an intervention on one arm. The somatosensory stimulus consists of 6 different intensities. So it is a within-subject design with 3 factors: Time (T0 and T1), Arm (intervention, control) and Intensity (6 different intensities).</div><div><br></div><div>I'm interested in the MAIN effect of INTENSITY, and the INTERACTION effects (TIME x ARM) and (TIME x ARM x INTENSITY).</div><div>For the Interaction effect TIME x ARM, I first calculated difference-waves (postcontrol - precontrol and postintervention - preintervention) in Matlab en then applied the permutation test on these difference-waves to test whether the two arms are different. I hope this is correct?</div><div><br></div><div>But now my question, how to test the TIME x ARM x INTENSITY interaction? </div><div>One possibility might be to calculate an ANOVA F statistic for this interaction effect (so perform a full factorial ANOVA) and perform the permutation test? Is this justified?</div><div><br></div><div>Very much thanks,</div><div>Best wishes,</div><div>Emanuel</div><div><br></div><div><br></div><div><br></div><div><br></div> </div></div></div> </div></body>
</html>