DOC Fix the description of some features in load_diabetes #19366
Conversation
|
Improved description of features |
sklearn/datasets/descr/diabetes.rst
Outdated
| @@ -21,11 +21,11 @@ quantitative measure of disease progression one year after baseline. | |||
| - sex | |||
| - bmi body mass index | |||
| - bp average blood pressure | |||
| - s1 tc, T-Cells (a type of white blood cells) | |||
| - s1 tc, total serum cholesterol | |||
| - s2 ldl, low-density lipoproteins | |||
| - s3 hdl, high-density lipoproteins | |||
| - s4 tch, thyroid stimulating hormone | |||
There was a problem hiding this comment.
tch is consistent with actually being total cholesterol over HDL, as my comments in issue #18940 indicate
There was a problem hiding this comment.
meaning s4 == s1 / s3? i tried comparing the values but they're not exactly the same. some rows differ by quite a bit. i'm hoping this is due to rounding error from the original dataset.
sklearn/datasets/descr/diabetes.rst
Outdated
| - s2 ldl, low-density lipoproteins | ||
| - s3 hdl, high-density lipoproteins | ||
| - s4 tch, thyroid stimulating hormone | ||
| - s5 ltg, lamotrigine | ||
| - s5 ltg, serum concentration of lamotrigine |
There was a problem hiding this comment.
ltg is almost certainly not anything to do with lamotrigine, an epilepsy drug. In communication with one of the original authors, Hastie, he and Efron indicated that it's probably log of triglycerides. That's a relevant quantity, although log_10 (which Efron suggested) gives unreasonable overall values (assuming the usual units). Natural log is more plausible. I'd suggest "ltg, possibly log serum triglycerides level". See my original comment in issue #18940
| - s4 tch, thyroid stimulating hormone | ||
| - s5 ltg, lamotrigine | ||
| - s4 tch, total cholesterol / HDL | ||
| - s5 ltg, possibly log of serum triglycerides level |
There was a problem hiding this comment.
| - s5 ltg, possibly log of serum triglycerides level | |
| - s5 ltg, log of serum triglycerides level |
There was a problem hiding this comment.
Maybe we should add a warning to state that the reliability of the meaning of each feature is not as good as we would have liked because the documentation of the source dataset is not very explicit.
There was a problem hiding this comment.
I think a note in the API docs would be more appropriate
There was a problem hiding this comment.
I added a note to sklearn.datasets.load_diabetes on how the meaning of each feature might not be clear.
sklearn/datasets/_base.py
Outdated
| @@ -757,7 +757,9 @@ def load_digits(*, n_class=10, return_X_y=False, as_frame=False): | |||
|
|
|||
| @_deprecate_positional_args | |||
| def load_diabetes(*, return_X_y=False, as_frame=False): | |||
| """Load and return the diabetes dataset (regression). | |||
| """Load and return the diabetes dataset (regression). The meaning of each | |||
There was a problem hiding this comment.
Let's put the note after the table below using sphinx:
.. note::
The meaning of each feature (i.e. `feature_names`) might be unclear (especially for `ltg`)
as the documentation of the original dataset is not explicit. We provide information that
seems correct in regard with the scientific literature in this field of research.
There was a problem hiding this comment.
Thanks for showing how sphinx can be used! Updated.
|
Thanks @hongshaoyang |
…rn#19366) Co-authored-by: Guillaume Lemaitre <g.lemaitre58@gmail.com>
…rn#19366) Co-authored-by: Guillaume Lemaitre <g.lemaitre58@gmail.com>
Co-authored-by: Guillaume Lemaitre <g.lemaitre58@gmail.com>
Reference Issues/PRs
Closes #18940
What does this implement/fix? Explain your changes.
Any other comments?