Added graph showing intra-subject variability (COV) as a function of atrophy estimation error

[PaulBautin]

I have opened this PR to present code used to generate figures in issue #86.

DONE:

• Added graphs showing intra-subject variability (COV) as a function of atrophy estimation error
• save output dataframes in a csv files

Fixes #86

[PaulBautin]

The following graphs are briefly introduced in the supplementary material of the manuscript and are only used to prove some of are hypotheses wrong. Should they be kept and merged? @jcohenadad
(Note: the results represented here are not the the results that will be published)

The following graph shows the non-correlation of t1w outliers (red dots) and t2w outliers (green dots)

[jcohenadad]

Should they be kept and merged?

hum... good question... just to be sure: you're talking about the additional functions (error_function_of_intra_cov*), right? I think it doesn't hurt to keep them in the main branch. The functions are well documented, and pretty "standalone". What do you think?

[PaulBautin]

Indeed, I am talking about the additional functions (error_function_of_intra_cov*). I agree that it would not hurt to keep these new functions in the main branch -- even more so because they are optional when running csa_rescale_stat.py.

In last commit ae45092, I have removed the following feature because it has become obsolete (there are no more missing CSAs).

* Test removing subjects with missing CSA: Presently if CSA is not computed the vertebrae levels are removed from the dataframe, I tried to see what happens when the entire subject is removed. It seems to eliminate values with high COV and low error % improving the correlation between these two metrics.

[jcohenadad]

thanks! great addition