Discussion about analysis and visualization of results

[valosekj]

I would like to discuss the best possible and most effective way how to analyze data and how to visualize all the results from large multi-subject dataset.
Since, I will analyze 6 qMRI metrics across individual vertebral levels and for individual ROI/labels, it will result into huge amount of data:

• dataset has >250 subjects
• 6 qMRI metrics will be analyzed - 4 for DWI (FA, AD, MD, RD) and 2 for MT (MTR, MTsat)
• DWI and MT data cover C2-C5 area, it means 4 individual levels for every subject (C2, C3, C4, C5) if I use -perlevel flag
• so far, I defined 6 ROI/labels - spinal cord, white matter, gray matter, ventral, lateral and dorsal columns/funiculi

I have created draft of possible tables layout here in google doc. Please, feel free to add your comments and ideas.

[jcohenadad]

personally i don't like huge tables with a bunch of numbers. It is useless and scares people out. what you want with your study is: (i) show an overview of the metrics trends and (ii) provide numbers for people to use.

so, with that goals in mind, what i would do is:

• (i) produce sexy plot with a reduced number of metrics (e.g. aggregated c2-c3)
• (ii) link to a versioned csv (with provenance about data that was used, software, analysis script, etc.), that people can import in excel, python, matlab, etc.

[valosekj]

• (i) produce sexy plot with a reduced number of metrics (e.g. aggregated c2-c3)
• (ii) link to a versioned csv (with provenance about data that was used, software, analysis script, etc.), that people can import in excel, python, matlab, etc.

Thank you! You are right, huge tables could be confusing and difficult to read. The idea to share csv files with all this data sounds better.