Thanks again Donald. My understanding is that if what one desires is the comparison among only two betas (e.g., g1c1 > g1c2), then the one-sample t-test (on the contrast images of the two betas) will produce equivalent results to the paired-sample t-test and flexible factorial in which the two betas are entered separately. Is this correct? ------------------------------------------------------------------------------- Bob Spunt Postdoctoral Fellow Social Cognitive and Affective Neuroscience Labs Department of Psychology University of California, Los Angeles On Wed, Feb 29, 2012 at 2:26 PM, MCLAREN, Donald <[log in to unmask]>wrote: > Yes. In more general terms: > If you are comparing a single beta to 0, you will want to use a one-sample > t-test. This is irrespective if it represents condition1 versus implicit > baseline or condition1-condition2. It only matters if its a single beta (or > group of beta tested against 0 [e.g. (g1c1+g1c2+g1c3)/3]). In other words, > if the comparison is not contrasting betas of two or more conditions > against each other, then you need to collapse observations and use a > one-sample t-test. > If you are comparing the beta for group1 condition 1 to group2 condition1, > you will want a two-sample t-test. This is irrespective if it represents > condition1 versus implicit baseline or condition1-condition2. It only > matters if its a two betas (or group of betas are comparing group effects > of the same betas [e.g. (g1c1+g1c2+g1c3)/3 versus(g2c1+g2c2+g2c3)/3 ]). In > other words, if the comparison is not contrasting betas of two or more > conditions against each other, then you need to collapse observations and > use a two-sample t-test. > etc. > > > 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 Wed, Feb 29, 2012 at 5:05 PM, Bob Spunt <[log in to unmask]> wrote: > >> Thanks Donald. Just to clarify, I assume that when you say >> between-subject effects you mean both (a) the betas for each regressor >> (i.e., condition compared to the implicit baseline) (in this case use a >> one-sample t-test), and (b) main effects of group (in this case use >> two-sample t-test or ANOVA). Or use GLM_flex, of course. Is this correct? >> >> Thanks again, >> Bob >> >> >> On Wed, Feb 29, 2012 at 1:51 PM, MCLAREN, Donald < >> [log in to unmask]> wrote: >> >>> Bob, >>> >>> You are absolutely correct that you don't need the main effects in the >>> model; however, I have noticed small changes in the effects when comparing >>> including them versus not including them. Best I can tell -- their >>> inclusion/exclusion effects either the omnibus test used to decide which >>> voxels to include OR the REML estimation of the variance-covariance pattern. >>> >>> As an additional note, the main effects of between-subject factors are >>> invalid because the error term and degrees of freedom are incorrect in the >>> full and flexible factorial models. To properly estimate the >>> between-subject effects you need to remove the repeated observations and >>> use a one-sample t-test, two-sample t-test or ANOVA. Alternatively, >>> GLM_flex properly estimates all effects (between and within-subjects) in a >>> single model. >>> >>> One other note in the tutorial, the main effect of condition is actually >>> a linear trend contrast, rather than the main effect. The main effect >>> F-contrast should have N-1 rows where N is the number of levels. >>> >>> >>> 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 Wed, Feb 29, 2012 at 4:36 PM, Bob Spunt <[log in to unmask]> wrote: >>> >>>> Dear SPM experts, >>>> >>>> I have a quick question regarding deciding which effects to include in >>>> a flexible factorial model. Consider the one-group case discussed in >>>> the Gläscher and Gitelman tutorial, which discusses a 2x3 within-subjects >>>> design (Pages 4-6). My question regards when to include the main effects in >>>> the model in addition to the interaction and subject effects, since it >>>> seems that all effects (namely, the two main effects and their interaction) >>>> can be tested in a model which includes only the interaction and subject >>>> effects. If this is indeed the case, then why would one ever include all >>>> effects in the model? >>>> >>>> Any light-shedding is much appreciated. >>>> >>>> Cheers, >>>> Bob >>>> >>>> >>>> ------------------------------------------------------------------------------- >>>> Bob Spunt >>>> Postdoctoral Fellow >>>> Social Cognitive and Affective Neuroscience Labs >>>> Department of Psychology >>>> University of California, Los Angeles >>>> >>>> >>> >> >