Issue 11 #20
Issue 11 #20
Conversation
There was a problem hiding this comment.
@nbokulich, generally the tests would go in at the same time as the new functionality. Are these in another PR right now?
q2_intervention/plugin_setup.py
Outdated
| 'either state_1 or state_2, that subject will be dropped and ' | ||
| 'reported in standard output by default. Set duplicates="ignore" ' | ||
| 'reported in standard output by default. Set replicates="ignore" ' |
There was a problem hiding this comment.
should this be random instead of ignore?
q2_intervention/_intervention.py
Outdated
|
|
||
|
|
||
| def paired_differences(output_dir: str, metadata: qiime2.Metadata, | ||
| group_column: str, metric: str, state_column: str, | ||
| state_1: str, state_2: str, individual_id_column: str, | ||
| parametric: bool=False, palette: str='Set1', | ||
| drop_duplicates: bool=True, table: pd.DataFrame=None | ||
| drop_replicates: str='error', table: pd.DataFrame=None |
There was a problem hiding this comment.
I don't think this should be called drop_replicates - maybe replicate_handling?
q2_intervention/_utilities.py
Outdated
| @@ -27,28 +27,40 @@ | |||
| TEMPLATES = pkg_resources.resource_filename('q2_intervention', 'assets') | |||
|
|
|||
|
|
|||
| def _check_inputs(state_1, state_2): | |||
There was a problem hiding this comment.
Could you rename this, maybe _validate_state_values, to be a little more self-documenting?
There was a problem hiding this comment.
I intend to validate more inputs in a future PR so will use _validate_input_values instead of _validate_state_values
|
thanks @gregcaporaso ! I added a unit test and revised per your suggestions. |
| results = [] | ||
| group_members = df[group_column] == group_value | ||
| group_md = df[group_members] | ||
| for individual_id in set(group_md[individual_id_column]): | ||
| result = [] | ||
| for state_value in state_values: | ||
| state_value = df[state_column].dtype.type(state_value) | ||
| individual_id = \ | ||
| df[individual_id_column].dtype.type(individual_id) | ||
| individual_id = df[individual_id_column].dtype.type(individual_id) | ||
| _state = df[state_column] == state_value | ||
| _ind = df[individual_id_column] == individual_id | ||
| individual_at_state_idx = group_md[_state & _ind].index | ||
| if len(individual_at_state_idx) > 1: | ||
| print("Multiple values for {0} {1} at {2} {3} ({4})".format( |
There was a problem hiding this comment.
These print statements aren't going to be very useful here, since with some interfaces users will see them, and with others they won't. In general, print statements shouldn't be in any QIIME actions. Instead, these types of warnings should be included in the visualization - for example, see the yellow warning box in this viz.
Since I know you don't have a lot of time to work on this before you leave, could you create an issue to replace these print statements with warnings in the visualization?
q2_longitudinal/_utilities.py
Outdated
| break | ||
| else: | ||
| elif replicate_handling == 'random': | ||
| individual_at_state_idx = [choice(individual_at_state_idx)] | ||
| elif len(individual_at_state_idx) == 0: |
There was a problem hiding this comment.
Instead of ending this with break, could you replace break with pass, and then put result.append(individual_at_state_idx[0]) in an else block?
q2_longitudinal/plugin_setup.py
Outdated
| 'detected, "error" causes method to fail; "drop" will discard all ' | ||
| 'subject IDs with replicate samples at either state_1 or state_2; ' | ||
| '"random" chooses one representative at random from among ' | ||
| 'replicates; "mean" and "median" compute average values across ' |
There was a problem hiding this comment.
mean and median aren't implemented, right? Can you remove those from this description?
|
thanks @gregcaporaso ! The latter two changes are made as suggested. The first change (about warning messages) has been turned into ticket #23 |
fixes #11
This does everything described in #11 with the exception of calculating mean and median values when replicates are found. This is a bit too complicated to implement at this stage for two reasons: (1) at this stage we are just grabbing the IDs to compare, and the pairwise difference/distance values are computed later; (2) this would behave very differently for each method — e.g., average alpha diversity is easy but how do we calculate average distance? If you think we should still support averaging replicates in a later release, we can leave this issue open or create a dedicated issue.