 |
 |
Dear Melanie, On Fri, 14 Jul 2017 13:22:53 +0100, Melanie Lafanechere <[log in to unmask]> wrote: >Dear experts, > >For my masters thesis I am doing SBM analysis, using a one-way ANOVA and I have a few questions regarding the ROI analysis and TFCE. > >1) A previous post says that there are two methods for doing ROI analyses; >A - an exported CSV file for all metrics that can be analysed using SPSS or >B - using CAT12's "Analyse ROI" option which uses the same design as my anova (including covariates etc). > >Using the data from the CSV files I get some significant findings using SPSS, but using the other method, nothing survives thresholding. > >1a) Does this mean that the values in the CSV files are uncorrected values? Yes. These are raw values where no potential nuisance variables are considered. >1b) Could you please clarify why one would use method A over method B to carry out the ROI analysis. Would method A be preferred if you have a very specific a priori hypothesis of where you would see a difference (i.e., highly localised/circumscribed to one region?). The CSV files existed first and represent the common way to make any additional analysis using an external software. Hence, you have to consider any additional corrections due to nuisance variables and also apply corrections for multiple comparisons. The "Analyze ROI" function is using the same GLM as the vertex- or voxel-wise analysis and also allows correction for multiple comparisons. >1c) Can I report the findings from the CSV files in my thesis, seeing as the other method doesn't yield any significant findings? > You still have to correct for multiple comparisons and if the results remain significant you can use the CSV results. >2) A previous post has also asked about the following TFCE output files, however I still don't really understand how to interpret these for my data, I've never really dealt with logarithms or this type of analysis before, so any further explanation would be really helpful! Am I right in thinking that the files without FDR or FWE appended to them provide uncorrected results. What is the difference between the files with Log_p and the others? Is the TFCE_ file providing corrected results? > >T_ raw T-values >TFCE_ raw TFCE values >TFCE_log_p log-scaled p-values for TFCE (uncorrected) >TFCE_log_pFDR log-scaled p-values for TFCE corrected using FDR >TFCE_log_pFWE log-scaled p-values for TFCE corrected using FWE >T_log_p same for T-values... >T_log_pFDR >T_log_pFWE > Best, Christian >Thanks in advance! > >Best wishes, >Melanie