As you noted, the con_* at the second level is identical. This is not surprising as the data you are asking about is the same. What is different is the variance terms between the models.

From my expierance the model with 4 conditions is wrong. The reason that it is wrong is that the error term represents the unexplained variance between your 4 categories (not the unexplained variance of all conditions). The model with 5 conditions has an error term that represents the unexplained variance of all the conditions.

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
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 Tue, May 24, 2011 at 7:01 AM, Henrike Hemingway <[log in to unmask]> wrote:
Dear SPMers,

I have question regarding the flexible factorial design as implemented in SPM8.

I have realized an experiment with 5 conditions for each subject (4 emotion conditions, 1 neutral condition). I was interested in the difference between the emotion and the neutral conditions.Therefore, I set up the 2nd level analysis using a flexible factorial design in SPM8 in two different ways. The main effect of subject and the main effect of condition were estimated in both of the design matrices.

In the first design matrix I specified the effect of condition with 5 levels, including the con images of all 5 conditions separately. The contrasts for each of the 4 conditions > neutral were computed on the 2nd level. For the second flexible factorial design I computed the differences between each of the 4 conditions > neutral on the first level and by this didn’t have to estimate difference contrasts on the 2nd level. Thus, I specified the effect of condition with only 4 levels each representing the difference of the condition > neutral and estimated the main effect for each of the four conditions.

The results for the contrasts representing the same difference (Condition > Neutral), however, were not similar across both designs as you can see in the attached screen shots. The model including the difference contrasts computed on the first level resulted in much higher t-values and larger number of voxels in the clusters of interest. After some research in the archives of the SPM mailing list I still have not found an explanation for the diverging results. Though I found some information about a comparable difference observed when comparing designs explicitly modeling the subject factor or not.

Does anyone know what exactly is causing the pronounced difference in both designs? When looking at the con images computed on the 2nd level they contain the exact same values in each voxel. Is one of the models practically wrong?

I would be glad if anyone could help me out with more detailed information or helpful links for further readings on that issue.

Thanks a lot.

Henrike