help > RE: intrepretation of [1 -1] contrast
Aug 11, 2017  06:08 PM
Alfonso Nieto-Castanon
McGovern Institute for Brain Research. MIT
RE: intrepretation of [1 -1] contrast
Hi Michael,

The corr* files in your second-level folders will contain simply the inverse-Fisher-transformed con_001 values (i.e. tanh(x) where x are the values in the con_001 file). Of course, these will be only particularly meaningful when, as you say, the original con_001 values represent some form of average connectivity (either a single condition or the average across several conditions). For all other cases simply use the original con_001 values for interpretation. Those always correspond to the effect-sizes that you are evaluating in your second-level analyses. For example in your case those will represent the average difference in Fisher-transformed correlation values between conditionA and conditionB. 

Best
Alfonso

Originally posted by ml90:
Hi,

I have an experiment with four conditions A, B, C and D and would like to check if seed-voxel connectivity is greater for condition A than condition B. I selected bivariate correlations at first level and at second level set Results.AllSubjects=1 and Results.between_conditions.contrast=[1 -1]. Is this correct? In the secondlevel folder I have corr_*.img files with the Fisher-transformed correlations. But how should I interpret these values? If the contrast was just [1], it would be the correlation coefficient between seed and voxel xyz for condition A, right? But what do the coefficients mean for the contrast [1 -1]?
Best regards,
Michael
reply reply with quote
Threaded View
TitleAuthorDate
intrepretation of [1 -1] contrastml90Aug 10, 2017
      RE: intrepretation of [1 -1] contrastAlfonso Nieto-CastanonAug 11, 2017