 |
 |
See inline responses below. Best Regards, Donald McLaren ================= D.G. McLaren University of Wisconsin - Madison Neuroscience Training Program Office: (608) 520-0586 ===================== 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 (608) 520-0586 or email. On Sun, Feb 14, 2010 at 2:48 AM, Bill Budd <[log in to unmask]>wrote: > Dear All, > > I'm a little confused about recent posts regarding the flexible factorial > model and the Glascher and Gitelman (2008) tutorial document Rachel posted > below. The recent posts from Darren and Donald indicate that you cannot use > a flexible factorial model to test between subject/group effects. I am wrong > in assuming that therefore the model/contrasts (Design 2 and 3) in the > tutorial document are incorrect? > Only the between subject contrasts are wrong, in my statistical opinion. The reason being is that there is a single error in SPM, whereas the majority of other programs have two error terms, one for within-subject error and one for between subject error. Some have argued that you can use partitioned error (the way SPM is set up), but most of the statisticians I have talked to prefer splitting the error term (the way SAS and SPSS are set up) as well as the point that your df is inflated in SPM because you have multiple observations for each subject in the between-subject measurements. For the interaction and between-subject, the within-subject error term is what should be used and is what SPM provides. > > Another recent post on the topic queried why a flexbile factorial model > should be used at all since both main effect of group and interactions can > be tested directly (without averaging over condition) using the full > factorial model. The tutorial document indicatethat including main > effects for subject and group in the design increases sensitivity. If so, is > it the case that subject and group factors should still be included in a > flexible factorial design, even though the error term is specified > incorrectly, but that only condition and group by condition interactions can > be tested? > I haven't compared the two models (flex versus factorial) to see if there are any differences. If the model is the same, as it should be, the only differences would be do to something being done behind the interface. As to the last question, you are correct. Include all three in the model (subject, group, condition) and the interaction if that is of interest. And only interpret the T-/F- tests for the condition or interaction terms. The beta estimates or contrasts for between-subject terms can be used, but the T-/F- test can't be interpreted. This allows you to accurately plot the response for each group-condition pairing without a separate model. This is possible because the estimates will not change because in the model you are averaging across all conditions in all subjects within a group which is the same as averaging across each condition in each subject and then averaging the result across all subjects within a group. Hope this clarifies your questions. > > Any clarification much appreciated! > > Cheers > -Bill > > > ------------------------------ > *From:* SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] *On > Behalf Of *Rachel Grossman > *Sent:* Sunday, 24 January 2010 6:51 PM > > *To:* [log in to unmask] > *Subject:* Re: [SPM] flexible facorial in spm5 > > Hi Jee Hye Seo, > > You may want to take a look on this excellent tutorial (see the attached). > > Rachel > > --- On *Sun, 1/24/10, Jee Hye Seo <[log in to unmask]>* wrote: > > > From: Jee Hye Seo <[log in to unmask]> > Subject: [SPM] flexible facorial in spm5 > To: [log in to unmask] > Date: Sunday, January 24, 2010, 12:49 AM > > Dear all, > I want to analysis usinf the flexible factorial in spm5. I have 3 groups > (n=20, n=18, n=22) and 3 conditions (rest, active1, active2). > I have some questions to all about flexible factorial in spm5. > > 1. what is the levels in condition? > 2. If I want to analysis the differences between 3 groups for active1 - > rest, how I construct flexible factorial in spm5? > how many factors? what is the levels? what is the design matrix? and so > on. > > I hope someone could help me with this questions. > All the best > > >