 |
 |
To my understanding, you do not need to model the control condition. So Vector A and B is enough. Jiansong > From: "Juanjo Lull (UPV)" <[log in to unmask]> > Reply-To: "Juanjo Lull (UPV)" <[log in to unmask]> > Date: Mon, 7 Feb 2005 12:44:46 +0100 > To: <[log in to unmask]> > Subject: Re: [SPM] What's differece between T-contrast and F-contrast? > > Dear Li Bei, > > > Li Bei wrote: > >> Dear Juan Jose Lull Noguera : >> >> Could you help me again? >> >> My experiment is designed like this: >> CACBCACBCACB ,where C means control,A means >> stimulus 1,B means stimulus 2, and where are 10 scans >> for each character >> >> (A,B,C),so there are 120 scans total .now i want to >> get the active region of stimuli A and B. >> In fmri design, i specify 3 conditions ,for A,B,C. >> Vector for A is [10 50 90] >> vector for B is [30 70 110] >> vector for C is [0 40 80] >> duration:10 >> In "result",i defined T contrast to show active >> regions in condition 1(A),so i wrote weight vector for >> T contrast [1 0 -1],which means A - C,then i got these >> active regions. --- >> Question 1,could you please tell me whether procedures >> above are correct??? >> >> > I think you meant C = [0 20 40 60 80 100] instead of [0 40 80]. If this > is true then the procedures described are ok. > >> Question 2,what's the meaning of F-contrast with >> vecotor 1 0 0 ? >> >> > Hmm. I don't see a meaning to this contrast, because you don't compare A > to anything... > >> Question 3,is there any way to get active regions of A >> with F-contrast? >> >> > Yes: you can apply an F [1 -1] contrast. This will give you activated > and also deactivated regions. If you want activated regions only, then > you should apply a t-test (F tests find differences in 'any' direction > while t-tests distinguish between directions). > >> Question 4,what's the meaning of [1 2 3] in >> F-contrast? >> >> > This doesn't tell me anything. I think (I am not a statistician so this > could be incorrect, but this is what I remember) that it would find > voxels that have a greater estimated activation (signal level) in C > related to B and in B related to A. > >> Question 5,in 'Result',there is always a predefined >> T-contrast,that is,"effects of interest".what's the >> vector of this T-contrast??? >> >> >> > It isn't a t-contrast but an f-contrast. There are some predefined > contrasts in SPM when you estimate a model. If you hadn't defined your > control (or baseline) task then it would find areas activated for > stimulus A and also areas activated for stimulus B. As you have > explicitly defined your baseline condition, I think that it doesn't give > you important information. > >> Those question puzzled me for a long time,thank >> you for your help!! >> >> >> >> > Cheers > Juan J > >> Li Bei >> >> >> >> >> --- Juan Jose Lull Noguera <[log in to unmask]> >> 的正文： >> >> >>> Hi Li Bei, >>> I recommend you to read a mail from Will Penny >>> written two days ago. The >>> messsage name is just "[SPM]". >>> If you have two conditions, c1 and c2 and you want >>> to see whether c1 is greater >>> than c2 then you should 'build' a [1 -1] t-contrast. >>> However, if you want to >>> see whether c2>c1 then you would build a [-1 1] >>> t-contrast (provided c1 is the >>> first condition in the design matrix, c2 the >>> second). >>> If you build an F contrast like [1 -1] it will test >>> whether there are >>> differences between c1 and c2, that is, c1>c2 and >>> also c2>c1. That's why you >>> see differences between a t and an F contrast. >>> Cheers >>> Juan J >>> >>> >> >> >> ===== >> Bei Li >> Institute of Neuroinformatics >> Dalian University of Technology >> 2 Ling Gong Rd, Dalian 116023, China >> Email:[log in to unmask] >> >> _________________________________________________________ >> Do You Yahoo!? >> 注册世界一流品质的雅虎免费电邮 >> http://cn.rd.yahoo.com/mail_cn/tag/1g/*http://cn.mail.yahoo.com/ >> >>