<html><head><style type='text/css'>p { margin: 0; }</style></head><body><div style='font-family: Times New Roman; font-size: 12pt; color: #000000'><P>Hi Nenad,</P>
<P> </P>
<P>Thanks for your question. These details need to be added to the wiki. </P>
<P> </P>
<P>First; ft_regressconfound can be applied to timelock, freq, and source data. The estimation of regression coefficients (beta weights of the head position data) is performed separately for each channel and each latency, in the case of timelock data. Consequently, after compensation, the sensor level data cannot be used anymore for source modeling. <SPAN style="FONT-FAMILY: 'Times New Roman','serif'; FONT-SIZE: 12pt; mso-fareast-font-family: 'Times New Roman'; mso-ansi-language: EN-US; mso-fareast-language: AR-SA; mso-bidi-language: AR-SA">To employ the GLM based compensation on the source level, single trial estimates for the cortical locations of interest have to be made from the original sensor level data, preferably using a common spatial filter based on all trials. The beta weights are subsequently estimated for each cortical location and the variance in source amplitude over trials that is explained by the head movement is removed.</SPAN></P>
<P><SPAN style="FONT-FAMILY: 'Times New Roman','serif'; FONT-SIZE: 12pt; mso-fareast-font-family: 'Times New Roman'; mso-ansi-language: EN-US; mso-fareast-language: AR-SA; mso-bidi-language: AR-SA"></SPAN> </P>
<P><SPAN style="FONT-FAMILY: 'Times New Roman','serif'; FONT-SIZE: 12pt; mso-fareast-font-family: 'Times New Roman'; mso-ansi-language: EN-US; mso-fareast-language: AR-SA; mso-bidi-language: AR-SA"></SPAN>We therefore recommend to use ft_regressconfound as a final step prior to ft_XXXstatistics. This is indeed in contrast to the ft_rejectXXX functions. Furthermore, make sure you select the same trials in the headposition data as you have selected in your MEG data since you are going to fit the two of them.</P>
<P> </P>
<P>Second, circumcenter.m (on the wiki) is a helper function that calculates one position and orientation of the head. This saves some degrees of freedom (df=6) as compared to taking into account the x,y,z-coordinates of each coil separately (n=3) as regressors (df=9). When you want to also use the squares, cubes, and derivatives as regressors, this can save quite a bit of degrees. The 'circumcenter' is the geometrical center of the three coils.</P>
<P> </P>
<P>Hope this answered your questions. Best regards,</P>
<P> </P>
<P>Arjen</P>
<P>
<HR id=zwchr>
</P>
<BLOCKQUOTE style="BORDER-LEFT: rgb(16,16,255) 2px solid; PADDING-LEFT: 5px; MARGIN-LEFT: 5px"><B>Van: </B>"Nenad Polomac" <polomacnenad@gmail.com><BR><B>Aan: </B>fieldtrip@science.ru.nl<BR><B>Verzonden: </B>Woensdag 14 november 2012 19:33:51<BR><B>Onderwerp: </B>Re: [FieldTrip] correction for movement tool<BR><BR>Hello Arjen,
<DIV><BR></DIV>
<DIV>Thank you very much for your suggestion. I think it will work in my data. Only thing that concerns me in <A href="http://fieldtrip.fcdonders.nl/example/how_to_incorporate_head_movements_in_meg_analysis" target=_blank>this tutorial</A> is when in the analysis pipeline I should perform this circumcenter calculation? Because if I do that e.g. before artifact rejection, I will have different numbers of trials in timelock and confound structures. Since in timelock I already removed some trials due to jumps or muscle artifacts.</DIV>
<DIV>So what is your suggestion for this?</DIV>
<DIV><BR></DIV>
<DIV>Thank you in advance!</DIV>
<DIV><BR></DIV>
<DIV>Nenad</DIV><BR>_______________________________________________<BR>fieldtrip mailing list<BR>fieldtrip@donders.ru.nl<BR>http://mailman.science.ru.nl/mailman/listinfo/fieldtrip</BLOCKQUOTE><BR></div></body></html>