Hi Patrick,

When using bivariate correlation measures in your first-level anlayses, the beta values (effect sizes) resulting from these analyses will represent functional connectivity between your seed and target area (separately for each subject and condition) in terms of Fisher transformed correlation coefficients. These values are then entered into your second-level model and a new set of beta values (effect sizes) is estimated from these second-level analyses now representing population- or group- level estimates. These (latter) beta values (always unstandardized beta values) may represent many different things depending on your second-level design. For example, they may represent average Fisher-transformed correlation coefficients across one group of subjects (e.g. in a one-sample t-test), average differences in Fisher-transformed correlation coefficients between two groups of subjects (e.g. in a two-sample t-test), expected increase in Fisher-trasnformed correlation coefficients for each unit-increase in some covariate (e.g. in a regression analysis), etc. When reporting the results of your second-level analyses it is always a good idea to, in addition to reporting the location of any significant area, also report the average effect size within those areas. In CONN you can do that using the 'display values' buttons which will show you a set of error bars representing the effect sizes within each cluster (or you can use the 'import values' buttons for creating other custom plots using the Calculator tool)

Hope this helps clarify
Alfonso


Originally posted by Patrick McConnell: