 |
 |
On Mon, Mar 26, 2012 at 8:35 AM, John Gelburg <[log in to unmask]>wrote: > Thank you Donald for your answer. Two follow-ups that I am missing: > > 1. As far as I understand, your approach does not try to compensate if for > one condition in total I have more trials than for the other. Wouldn't it > create I bias? Consider the case that in one condition I have in total > let's say two times more trial, but in case that trials evenly distributed > across session the contrast would look like: [0.5 -0.5 0.5 -0.5]. > If you have null hypothesis c1=c2. The weighting is based on the number of trials on the left side of the equation for the left sided weights and the weighting on the right sides is based on the number trials on the right side. Once you have the weightings, then subtract c2 from c1. The more trials you have the closer the estimate will be to the true response. > > 2. You suppose that two sessions are equivalent, but if for some reason a > subject was more alert in one of the sessions, this might create a bias > towards the condition, which has more trials in this session. Isn't it? > This would be an instance of when you might not want to weight the conditions based on the number of trials. > > Thanks again > > > On Sun, Mar 25, 2012 at 11:54 PM, MCLAREN, Donald < > [log in to unmask]> wrote: > >> It depends on what inference you want to draw, but generally I would >> recommend that you create a weighted average of the sessions: >> c1s1 -- 30 trials >> c1s2 -- 15 trials >> c2s1 -- 14 trials >> c2s2 -- 16 trials >> contrast=[30/45 -14/30 15/45 -16/30] >> >> In this, you are comparing the average response of c1 to c2. >> >> This is exactly what is done with gPPI for event-related designs. >> >> Best Regards, Donald McLaren >> ================= >> D.G. McLaren, Ph.D. >> Postdoctoral Research Fellow, GRECC, Bedford VA >> Research Fellow, Department of Neurology, Massachusetts General Hospital >> and >> Harvard Medical School >> Website: http://www.martinos.org/~mclaren >> Office: (773) 406-2464 >> ===================== >> This e-mail contains CONFIDENTIAL INFORMATION which may contain PROTECTED >> HEALTHCARE INFORMATION and may also be LEGALLY PRIVILEGED and which is >> intended only for the use of the individual or entity named above. If the >> reader of the e-mail is not the intended recipient or the employee or >> agent >> responsible for delivering it to the intended recipient, you are hereby >> notified that you are in possession of confidential and privileged >> information. Any unauthorized use, disclosure, copying or the taking of >> any >> action in reliance on the contents of this information is strictly >> prohibited and may be unlawful. If you have received this e-mail >> unintentionally, please immediately notify the sender via telephone at >> (773) >> 406-2464 or email. >> >> >> >> >> On Sun, Mar 25, 2012 at 5:41 PM, John Gelburg <[log in to unmask]>wrote: >> >>> Hi all, >>> >>> Any help would be very appreciated. >>> >>> I have two conditions design and two sessions. The number of trials >>> per condition is defined by subject's response in the scanner. So, I >>> end up with different number of repetitions per condition. >>> >>> 1. If in first session I have more trials of condition 1 and in the >>> second session I have more trials of condition 2, would it be >>> incorrect to define a general t-contrast [1 -1 1 -1]? Clearly, I can >>> equalize and take for each session the minimal number of trials, but >>> then I lose some data. >>> >>> 2. Does it make sense to put the weights on contrast definition? How >>> would look a t-contrast if let's say in cond1 I have 10 trials and in >>> cond2 only 5? >>> >>> Thanks a lot, >>> John >>> >> >> >