Profile format changed to be less tall and more wide, like mAFQ #279
Conversation
| @@ -192,7 +192,7 @@ def test_AFQ_data_waypoint(): | |||
| 'sub-01_sess-01_dwi_space-RASMM_model-DTI_desc-det-AFQ-clean_tractography_idx.json')) # noqa | |||
|
|
|||
| tract_profiles = pd.read_csv(myafq.tract_profiles[0]) | |||
| assert tract_profiles.shape == (800, 5) | |||
| assert tract_profiles.shape == (400, 5) | |||
There was a problem hiding this comment.
Hmm. Why are there only half the numbers here?
There was a problem hiding this comment.
Two columns were removed: scalar and profile. scalar was the label (ie "dti_fa") and profile was the number. Now, instead, there are two columns of numbers: "dti_md", "dti_fa". This means it is only half the length. For each node, there is only one row instead of one for each scalar.
There was a problem hiding this comment.
In general, I am not sure that I like this change. Maybe we should change the way things get converted from mAFQ into AFQ-Insight instead? WDYT @richford? Is there any particular reason to prefer long over wide or vice versa?
|
No reason other than the appeal to tidy data. That's just the format that we used for AFQ-Browser and AFQ-Insight. Inside of AFQ-insight, we want the feature matrix to be wide (n x p, where n in the number of subjects and p is the number of features), but I can definitely do that conversion inside of AFQ-Insight. I don't remember if we had good reasons to prefer the wide format inside of AFQ-Browser other than the appeal to tidy data. |
|
Sorry for the lingo. I'm referring to this definition of tidy. Hadley gets extra points for the Tolstoy quote. |
|
OK - sounds good. Let's go with the wide format proposed here (and already implemented in AFQ-Browser and AFQ-Insight). Only still need to understand the changes introduced in the tests, but otherwise, LGTM. |
|
With that question, I believe this is ready to merge? Is it rebased on master? |
|
I just rebased it. I believe it can merge when these checks complete. |
Instead of a scalar column with the names "dti_fa", "dti_md" and a profiles column with their corresponding values, each scalar gets its own column. closes #277