sdm-help-list
sdm-help-list > Interpreting the results of a linear model analysis
Feb 22, 2021 12:02 AM | Andrew Northrop
Interpreting the results of a linear model analysis
Hi all,
I recently performed a linear model analysis using the YBOCS variable in the tutorial of SDM-PSI v6.21 and I'm wondering how I should interpret the results. I extracted the values from one of the peaks that survived a FWE correction (see attached) and I have a few questions about what the different values mean:
1. It seems like the main measure of effect for each peak is the SDM-Z value of the effect size. What does this value represent? Do positive Z values mean that the region is more activated for higher values of YBOCS (and negative Z values that the regions is more activated for lower values of YBOCS)?
2. Is there an output value that gives the slope of the linear model for a given peak?
3. How should I interpret the heterogeneity statistics of each peak(I^2 and Q^2)? In other types of meta-analyses, it is conventional to only perform meta-regressions for a particular outcome if these statistics exceed a certain value (e.g. I^2 = 50). Is there a similar threshold used for meta-analyses in SDM-PSI? What range of values should I expect for my I^2 and Q^2 values?
I recently performed a linear model analysis using the YBOCS variable in the tutorial of SDM-PSI v6.21 and I'm wondering how I should interpret the results. I extracted the values from one of the peaks that survived a FWE correction (see attached) and I have a few questions about what the different values mean:
1. It seems like the main measure of effect for each peak is the SDM-Z value of the effect size. What does this value represent? Do positive Z values mean that the region is more activated for higher values of YBOCS (and negative Z values that the regions is more activated for lower values of YBOCS)?
2. Is there an output value that gives the slope of the linear model for a given peak?
3. How should I interpret the heterogeneity statistics of each peak(I^2 and Q^2)? In other types of meta-analyses, it is conventional to only perform meta-regressions for a particular outcome if these statistics exceed a certain value (e.g. I^2 = 50). Is there a similar threshold used for meta-analyses in SDM-PSI? What range of values should I expect for my I^2 and Q^2 values?