Skip to content

quantile_binning

Transformations for normalizing MEDS datasets, across both categorical and continuous dimensions.

DO NOT RUN THIS WITH PARALLELISM

add_custom_quantiles_column(code_metadata, custom_quantiles)

Add a custom_quantiles column to code_metadata DataFrame based on provided dictionary.

Parameters:

Name Type Description Default
code_metadata DataFrame

A Polars DataFrame containing code information

 required
custom_quantiles dict

A dictionary mapping codes to their custom quantile values

 required

Returns:

Type Description
DataFrame

A DataFrame with an added custom_quantiles column

Examples:

import polars as pl code_metadata = pl.DataFrame({ … “code”: [“lab//A”, “lab//B”, “lab//C”], … }) custom_quantiles = { … “lab//A”: {“values/quantile/0.5”: 1.5}, … “lab//B”: {“values/quantile/0.5”: 2.5} … } result = add_custom_quantiles_column(code_metadata, custom_quantiles) result shape: (3, 2) ┌────────┬──────────────────┐ │ code ┆ custom_quantiles │ │ — ┆ — │ │ str ┆ struct[1] │ ╞════════╪══════════════════╡ │ lab//A ┆ {1.5} │ │ lab//B ┆ {2.5} │ │ lab//C ┆ null │ └────────┴──────────────────┘

Test with empty custom_quantiles

result = add_custom_quantiles_column(code_metadata, {}) result shape: (3, 2) ┌────────┬──────────────────┐ │ code ┆ custom_quantiles │ │ — ┆ — │ │ str ┆ null │ ╞════════╪══════════════════╡ │ lab//A ┆ null │ │ lab//B ┆ null │ │ lab//C ┆ null │ └────────┴──────────────────┘

Source code in meds_torch/utils/quantile_binning.py 
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
def add_custom_quantiles_column(code_metadata: pl.DataFrame, custom_quantiles: dict) -> pl.DataFrame:
    """Add a custom_quantiles column to code_metadata DataFrame based on provided
    dictionary.

    Args:
        code_metadata: A Polars DataFrame containing code information
        custom_quantiles: A dictionary mapping codes to their custom quantile values

    Returns:
        A DataFrame with an added custom_quantiles column

    Examples:
    >>> import polars as pl
    >>> code_metadata = pl.DataFrame({
    ...     "code": ["lab//A", "lab//B", "lab//C"],
    ... })
    >>> custom_quantiles = {
    ...     "lab//A": {"values/quantile/0.5": 1.5},
    ...     "lab//B": {"values/quantile/0.5": 2.5}
    ... }
    >>> result = add_custom_quantiles_column(code_metadata, custom_quantiles)
    >>> result
    shape: (3, 2)
    ┌────────┬──────────────────┐
    │ code   ┆ custom_quantiles │
    │ ---    ┆ ---              │
    │ str    ┆ struct[1]        │
    ╞════════╪══════════════════╡
    │ lab//A ┆ {1.5}            │
    │ lab//B ┆ {2.5}            │
    │ lab//C ┆ null             │
    └────────┴──────────────────┘
    >>> # Test with empty custom_quantiles
    >>> result = add_custom_quantiles_column(code_metadata, {})
    >>> result
    shape: (3, 2)
    ┌────────┬──────────────────┐
    │ code   ┆ custom_quantiles │
    │ ---    ┆ ---              │
    │ str    ┆ null             │
    ╞════════╪══════════════════╡
    │ lab//A ┆ null             │
    │ lab//B ┆ null             │
    │ lab//C ┆ null             │
    └────────┴──────────────────┘
    """
    # Convert custom_quantiles dict to a Polars Series
    if isinstance(custom_quantiles, DictConfig):
        custom_quantiles = OmegaConf.to_container(custom_quantiles)
    custom_quantiles_series = pl.Series(
        name="custom_quantiles",
        values=[custom_quantiles.get(code) if code is not None else None for code in code_metadata["code"]],
    )

    # Add the custom_quantiles column to code_metadata
    return code_metadata.with_columns(custom_quantiles_series)

convert_metadata_codes_to_discrete_quantiles(code_metadata, custom_quantiles)

Converts the numeric values in a MEDS dataset to discrete quantiles that are added to the code name.

Returns:

Type Description
DataFrame
  • A new DataFrame with the quantile values added to the code name.
DataFrame
  • A new DataFrame with the metadata for the quantile codes.

Examples:

from datetime import datetime code_metadata = pl.DataFrame( … { … “code”: [“lab//A”, “lab//C”, “dx//B”, “dx//E”, “lab//F”, “dx//D”], … “values/quantiles”: [ # [[-3,-1,1,3], [-3,-1,1,3], [], [-3,-1,1,3], [-3,-1,1,3], [-3,-1,1,3]], … {“values/quantile/0.2”: -3, “values/quantile/0.4”: -1, … “values/quantile/0.6”: 1, “values/quantile/0.8”: 3}, … {“values/quantile/0.2”: -3, “values/quantile/0.4”: -1, … “values/quantile/0.6”: 1, “values/quantile/0.8”: 3}, … {“values/quantile/0.2”: None, “values/quantile/0.4”: None, … “values/quantile/0.6”: None, “values/quantile/0.8”: None}, … {“values/quantile/0.2”: -3, “values/quantile/0.4”: -1, … “values/quantile/0.6”: 1, “values/quantile/0.8”: 3}, … {“values/quantile/0.2”: -3, “values/quantile/0.4”: -1, … “values/quantile/0.6”: 1, “values/quantile/0.8”: 3}, … {“values/quantile/0.2”: None, “values/quantile/0.4”: None, … “values/quantile/0.6”: None, “values/quantile/0.8”: None}, … ], … }, … schema = { … “code”: pl.Utf8, … “values/quantiles”: pl.Struct([ … pl.Field(“values/quantile/0.2”, pl.Float64), … pl.Field(“values/quantile/0.4”, pl.Float64), … pl.Field(“values/quantile/0.6”, pl.Float64), … pl.Field(“values/quantile/0.8”, pl.Float64), … ]), # pl.List(pl.Float64), … }, … ) custom_quantiles = {“lab//C”: {“values/quantile/0.5”: 0}} quantile_code_metadata = convert_metadata_codes_to_discrete_quantiles( … code_metadata, custom_quantiles) quantile_code_metadata.sort(“code”) shape: (25, 2) ┌──────────────┬───────────────────────┐ │ code ┆ values/quantiles │ │ — ┆ — │ │ str ┆ struct[4] │ ╞══════════════╪═══════════════════════╡ │ dx//B ┆ {null,null,null,null} │ │ dx//D ┆ {null,null,null,null} │ │ dx//E ┆ {-3.0,-1.0,1.0,3.0} │ │ dx//E//_Q_1 ┆ {-3.0,-1.0,1.0,3.0} │ │ dx//E//_Q_2 ┆ {-3.0,-1.0,1.0,3.0} │ │ … ┆ … │ │ lab//F//_Q_1 ┆ {-3.0,-1.0,1.0,3.0} │ │ lab//F//_Q_2 ┆ {-3.0,-1.0,1.0,3.0} │ │ lab//F//_Q_3 ┆ {-3.0,-1.0,1.0,3.0} │ │ lab//F//_Q_4 ┆ {-3.0,-1.0,1.0,3.0} │ │ lab//F//_Q_5 ┆ {-3.0,-1.0,1.0,3.0} │ └──────────────┴───────────────────────┘

Source code in meds_torch/utils/quantile_binning.py 
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
def convert_metadata_codes_to_discrete_quantiles(
    code_metadata: pl.DataFrame, custom_quantiles
) -> pl.DataFrame:
    """Converts the numeric values in a MEDS dataset to discrete quantiles that are
    added to the code name.

    Returns:
        - A new DataFrame with the quantile values added to the code name.
        - A new DataFrame with the metadata for the quantile codes.

    Examples:
    >>> from datetime import datetime
    >>> code_metadata = pl.DataFrame(
    ...     {
    ...         "code": ["lab//A", "lab//C", "dx//B", "dx//E", "lab//F", "dx//D"],
    ...         "values/quantiles": [ # [[-3,-1,1,3], [-3,-1,1,3], [], [-3,-1,1,3], [-3,-1,1,3], [-3,-1,1,3]],
    ...             {"values/quantile/0.2": -3, "values/quantile/0.4": -1,
    ...                 "values/quantile/0.6": 1, "values/quantile/0.8": 3},
    ...             {"values/quantile/0.2": -3, "values/quantile/0.4": -1,
    ...                 "values/quantile/0.6": 1, "values/quantile/0.8": 3},
    ...             {"values/quantile/0.2": None, "values/quantile/0.4": None,
    ...                 "values/quantile/0.6": None, "values/quantile/0.8": None},
    ...             {"values/quantile/0.2": -3, "values/quantile/0.4": -1,
    ...                 "values/quantile/0.6": 1, "values/quantile/0.8": 3},
    ...             {"values/quantile/0.2": -3, "values/quantile/0.4": -1,
    ...                 "values/quantile/0.6": 1, "values/quantile/0.8": 3},
    ...             {"values/quantile/0.2": None, "values/quantile/0.4": None,
    ...                 "values/quantile/0.6": None, "values/quantile/0.8": None},
    ...         ],
    ...     },
    ...     schema = {
    ...         "code": pl.Utf8,
    ...         "values/quantiles": pl.Struct([
    ...             pl.Field("values/quantile/0.2", pl.Float64),
    ...             pl.Field("values/quantile/0.4", pl.Float64),
    ...             pl.Field("values/quantile/0.6", pl.Float64),
    ...             pl.Field("values/quantile/0.8", pl.Float64),
    ...         ]), # pl.List(pl.Float64),
    ...     },
    ... )
    >>> custom_quantiles = {"lab//C": {"values/quantile/0.5": 0}}
    >>> quantile_code_metadata = convert_metadata_codes_to_discrete_quantiles(
    ...    code_metadata, custom_quantiles)
    >>> quantile_code_metadata.sort("code")
    shape: (25, 2)
    ┌──────────────┬───────────────────────┐
    │ code         ┆ values/quantiles      │
    │ ---          ┆ ---                   │
    │ str          ┆ struct[4]             │
    ╞══════════════╪═══════════════════════╡
    │ dx//B        ┆ {null,null,null,null} │
    │ dx//D        ┆ {null,null,null,null} │
    │ dx//E        ┆ {-3.0,-1.0,1.0,3.0}   │
    │ dx//E//_Q_1  ┆ {-3.0,-1.0,1.0,3.0}   │
    │ dx//E//_Q_2  ┆ {-3.0,-1.0,1.0,3.0}   │
    │ …            ┆ …                     │
    │ lab//F//_Q_1 ┆ {-3.0,-1.0,1.0,3.0}   │
    │ lab//F//_Q_2 ┆ {-3.0,-1.0,1.0,3.0}   │
    │ lab//F//_Q_3 ┆ {-3.0,-1.0,1.0,3.0}   │
    │ lab//F//_Q_4 ┆ {-3.0,-1.0,1.0,3.0}   │
    │ lab//F//_Q_5 ┆ {-3.0,-1.0,1.0,3.0}   │
    └──────────────┴───────────────────────┘
    """
    # Step 1: Add custom_quantiles column to code_metadata
    final_metadata_columns = code_metadata.columns
    code_metadata = add_custom_quantiles_column(code_metadata, custom_quantiles)

    # Step 2: Generate quantile_code_metadata
    quantile_code_metadata = generate_quantile_code_metadata(code_metadata)

    return quantile_code_metadata.select(final_metadata_columns)

convert_to_discrete_quantiles(meds_data, code_metadata, custom_quantiles)

Converts the numeric values in a MEDS dataset to discrete quantiles that are added to the code name.

Returns:

Type Description
DataFrame
  • A new DataFrame with the quantile values added to the code name.
DataFrame
  • A new DataFrame with the metadata for the quantile codes.

Examples:

from datetime import datetime MEDS_df = pl.DataFrame( … { … “subject_id”: [1, 1, 1, 2, 2, 2, 3], … “time”: [ … datetime(2021, 1, 1), … datetime(2021, 1, 1), … datetime(2021, 1, 2), … datetime(2022, 10, 2), … datetime(2022, 10, 2), … datetime(2022, 10, 2), … datetime(2022, 10, 2), … ], … “code”: [“lab//A”, “lab//C”, “dx//B”, “lab//A”, “dx//D”, “lab//C”, “lab//F”], … “numeric_value”: [1, 3, None, 3, None, None, None], … }, … schema = { … “subject_id”: pl.UInt32, … “time”: pl.Datetime, … “code”: pl.Utf8, … “numeric_value”: pl.Float64, … }, … ) code_metadata = pl.DataFrame( … { … “code”: [“lab//A”, “lab//C”, “dx//B”, “dx//E”, “lab//F”, “dx//D”], … “values/quantiles”: [ # [[-3,-1,1,3], [-3,-1,1,3], [], [-3,-1,1,3], [-3,-1,1,3], [-3,-1,1,3]], … {“values/quantile/0.2”: -3, “values/quantile/0.4”: -1, … “values/quantile/0.6”: 1, “values/quantile/0.8”: 3}, … {“values/quantile/0.2”: -3, “values/quantile/0.4”: -1, … “values/quantile/0.6”: 1, “values/quantile/0.8”: 3}, … {“values/quantile/0.2”: None, “values/quantile/0.4”: None, … “values/quantile/0.6”: None, “values/quantile/0.8”: None}, … {“values/quantile/0.2”: -3, “values/quantile/0.4”: -1, … “values/quantile/0.6”: 1, “values/quantile/0.8”: 3}, … {“values/quantile/0.2”: -3, “values/quantile/0.4”: -1, … “values/quantile/0.6”: 1, “values/quantile/0.8”: 3}, … {“values/quantile/0.2”: -3, “values/quantile/0.4”: -1, … “values/quantile/0.6”: 1, “values/quantile/0.8”: 3}, … ], … }, … schema = { … “code”: pl.Utf8, … “values/quantiles”: pl.Struct([ … pl.Field(“values/quantile/0.2”, pl.Float64), … pl.Field(“values/quantile/0.4”, pl.Float64), … pl.Field(“values/quantile/0.6”, pl.Float64), … pl.Field(“values/quantile/0.8”, pl.Float64), … ]), # pl.List(pl.Float64), … }, … ) custom_quantiles = {“lab//C”: {“values/quantile/0.5”: 0}} result = convert_to_discrete_quantiles( … MEDS_df, code_metadata, custom_quantiles) result.sort(“subject_id”, “time”, “code”) shape: (7, 4) ┌────────────┬─────────────────────┬──────────────┬───────────────┐ │ subject_id ┆ time ┆ code ┆ numeric_value │ │ — ┆ — ┆ — ┆ — │ │ u32 ┆ datetime[μs] ┆ str ┆ f64 │ ╞════════════╪═════════════════════╪══════════════╪═══════════════╡ │ 1 ┆ 2021-01-01 00:00:00 ┆ lab//A//_Q_3 ┆ 1.0 │ │ 1 ┆ 2021-01-01 00:00:00 ┆ lab//C//_Q_2 ┆ 3.0 │ │ 1 ┆ 2021-01-02 00:00:00 ┆ dx//B ┆ null │ │ 2 ┆ 2022-10-02 00:00:00 ┆ dx//D ┆ null │ │ 2 ┆ 2022-10-02 00:00:00 ┆ lab//A//_Q_4 ┆ 3.0 │ │ 2 ┆ 2022-10-02 00:00:00 ┆ lab//C ┆ null │ │ 3 ┆ 2022-10-02 00:00:00 ┆ lab//F ┆ null │ └────────────┴─────────────────────┴──────────────┴───────────────┘ custom_quantiles = {“lab//A”: {“values/quantile/0.5”: 3}} convert_to_discrete_quantiles(MEDS_df, code_metadata, custom_quantiles … ).sort(“subject_id”, “time”, “code”) shape: (7, 4) ┌────────────┬─────────────────────┬──────────────┬───────────────┐ │ subject_id ┆ time ┆ code ┆ numeric_value │ │ — ┆ — ┆ — ┆ — │ │ u32 ┆ datetime[μs] ┆ str ┆ f64 │ ╞════════════╪═════════════════════╪══════════════╪═══════════════╡ │ 1 ┆ 2021-01-01 00:00:00 ┆ lab//A//_Q_1 ┆ 1.0 │ │ 1 ┆ 2021-01-01 00:00:00 ┆ lab//C//_Q_4 ┆ 3.0 │ │ 1 ┆ 2021-01-02 00:00:00 ┆ dx//B ┆ null │ │ 2 ┆ 2022-10-02 00:00:00 ┆ dx//D ┆ null │ │ 2 ┆ 2022-10-02 00:00:00 ┆ lab//A//_Q_1 ┆ 3.0 │ │ 2 ┆ 2022-10-02 00:00:00 ┆ lab//C ┆ null │ │ 3 ┆ 2022-10-02 00:00:00 ┆ lab//F ┆ null │ └────────────┴─────────────────────┴──────────────┴───────────────┘

Source code in meds_torch/utils/quantile_binning.py 
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
def convert_to_discrete_quantiles(
    meds_data: pl.DataFrame, code_metadata: pl.DataFrame, custom_quantiles
) -> pl.DataFrame:
    """Converts the numeric values in a MEDS dataset to discrete quantiles that are
    added to the code name.

    Returns:
        - A new DataFrame with the quantile values added to the code name.
        - A new DataFrame with the metadata for the quantile codes.

    Examples:
    >>> from datetime import datetime
    >>> MEDS_df = pl.DataFrame(
    ...     {
    ...         "subject_id": [1, 1, 1, 2, 2, 2, 3],
    ...         "time": [
    ...             datetime(2021, 1, 1),
    ...             datetime(2021, 1, 1),
    ...             datetime(2021, 1, 2),
    ...             datetime(2022, 10, 2),
    ...             datetime(2022, 10, 2),
    ...             datetime(2022, 10, 2),
    ...             datetime(2022, 10, 2),
    ...         ],
    ...         "code": ["lab//A", "lab//C", "dx//B", "lab//A", "dx//D", "lab//C", "lab//F"],
    ...         "numeric_value": [1, 3, None, 3, None, None, None],
    ...     },
    ...     schema = {
    ...         "subject_id": pl.UInt32,
    ...         "time": pl.Datetime,
    ...         "code": pl.Utf8,
    ...         "numeric_value": pl.Float64,
    ...    },
    ... )
    >>> code_metadata = pl.DataFrame(
    ...     {
    ...         "code": ["lab//A", "lab//C", "dx//B", "dx//E", "lab//F", "dx//D"],
    ...         "values/quantiles": [ # [[-3,-1,1,3], [-3,-1,1,3], [], [-3,-1,1,3], [-3,-1,1,3], [-3,-1,1,3]],
    ...             {"values/quantile/0.2": -3, "values/quantile/0.4": -1,
    ...                 "values/quantile/0.6": 1, "values/quantile/0.8": 3},
    ...             {"values/quantile/0.2": -3, "values/quantile/0.4": -1,
    ...                 "values/quantile/0.6": 1, "values/quantile/0.8": 3},
    ...             {"values/quantile/0.2": None, "values/quantile/0.4": None,
    ...                 "values/quantile/0.6": None, "values/quantile/0.8": None},
    ...             {"values/quantile/0.2": -3, "values/quantile/0.4": -1,
    ...                 "values/quantile/0.6": 1, "values/quantile/0.8": 3},
    ...             {"values/quantile/0.2": -3, "values/quantile/0.4": -1,
    ...                 "values/quantile/0.6": 1, "values/quantile/0.8": 3},
    ...             {"values/quantile/0.2": -3, "values/quantile/0.4": -1,
    ...                 "values/quantile/0.6": 1, "values/quantile/0.8": 3},
    ...         ],
    ...     },
    ...     schema = {
    ...         "code": pl.Utf8,
    ...         "values/quantiles": pl.Struct([
    ...             pl.Field("values/quantile/0.2", pl.Float64),
    ...             pl.Field("values/quantile/0.4", pl.Float64),
    ...             pl.Field("values/quantile/0.6", pl.Float64),
    ...             pl.Field("values/quantile/0.8", pl.Float64),
    ...         ]), # pl.List(pl.Float64),
    ...     },
    ... )
    >>> custom_quantiles = {"lab//C": {"values/quantile/0.5": 0}}
    >>> result = convert_to_discrete_quantiles(
    ...    MEDS_df, code_metadata, custom_quantiles)
    >>> result.sort("subject_id", "time", "code")
    shape: (7, 4)
    ┌────────────┬─────────────────────┬──────────────┬───────────────┐
    │ subject_id ┆ time                ┆ code         ┆ numeric_value │
    │ ---        ┆ ---                 ┆ ---          ┆ ---           │
    │ u32        ┆ datetime[μs]        ┆ str          ┆ f64           │
    ╞════════════╪═════════════════════╪══════════════╪═══════════════╡
    │ 1          ┆ 2021-01-01 00:00:00 ┆ lab//A//_Q_3 ┆ 1.0           │
    │ 1          ┆ 2021-01-01 00:00:00 ┆ lab//C//_Q_2 ┆ 3.0           │
    │ 1          ┆ 2021-01-02 00:00:00 ┆ dx//B        ┆ null          │
    │ 2          ┆ 2022-10-02 00:00:00 ┆ dx//D        ┆ null          │
    │ 2          ┆ 2022-10-02 00:00:00 ┆ lab//A//_Q_4 ┆ 3.0           │
    │ 2          ┆ 2022-10-02 00:00:00 ┆ lab//C       ┆ null          │
    │ 3          ┆ 2022-10-02 00:00:00 ┆ lab//F       ┆ null          │
    └────────────┴─────────────────────┴──────────────┴───────────────┘
    >>> custom_quantiles = {"lab//A": {"values/quantile/0.5": 3}}
    >>> convert_to_discrete_quantiles(MEDS_df, code_metadata, custom_quantiles
    ...     ).sort("subject_id", "time", "code")
    shape: (7, 4)
    ┌────────────┬─────────────────────┬──────────────┬───────────────┐
    │ subject_id ┆ time                ┆ code         ┆ numeric_value │
    │ ---        ┆ ---                 ┆ ---          ┆ ---           │
    │ u32        ┆ datetime[μs]        ┆ str          ┆ f64           │
    ╞════════════╪═════════════════════╪══════════════╪═══════════════╡
    │ 1          ┆ 2021-01-01 00:00:00 ┆ lab//A//_Q_1 ┆ 1.0           │
    │ 1          ┆ 2021-01-01 00:00:00 ┆ lab//C//_Q_4 ┆ 3.0           │
    │ 1          ┆ 2021-01-02 00:00:00 ┆ dx//B        ┆ null          │
    │ 2          ┆ 2022-10-02 00:00:00 ┆ dx//D        ┆ null          │
    │ 2          ┆ 2022-10-02 00:00:00 ┆ lab//A//_Q_1 ┆ 3.0           │
    │ 2          ┆ 2022-10-02 00:00:00 ┆ lab//C       ┆ null          │
    │ 3          ┆ 2022-10-02 00:00:00 ┆ lab//F       ┆ null          │
    └────────────┴─────────────────────┴──────────────┴───────────────┘
    """
    # Step 1: Add custom_quantiles column to code_metadata
    code_metadata = add_custom_quantiles_column(code_metadata, custom_quantiles)

    # Step 2: Join meds_data with the updated code_metadata
    result = meds_data.join(code_metadata, on="code", how="left")

    # Step 3: Process quantiles, prioritizing custom_quantiles
    result = process_quantiles(result)

    return result

generate_quantile_code_metadata(code_metadata)

Modifies the code_metadata DataFrame to include quantile codes.

Parameters:

Name Type Description Default
code_metadata DataFrame

Current code metadata DataFrame

 required

Returns:

Type Description
DataFrame

pl.DataFrame: dataframe where we duplicate rows for each quantile code.

Examples:

from datetime import datetime code_metadata = pl.DataFrame( … { … “code”: [“lab//A”, “lab//C”, “dx//B”, “dx//E”, “lab//F”, “dx//D”], … “values/quantiles”: [ # [[-3,-1,1,3], [-3,-1,1,3], [], [-3,-1,1,3], [-3,-1,1,3], [-3,-1,1,3]], … {“values/quantile/0.2”: -3, “values/quantile/0.4”: -1, … “values/quantile/0.6”: 1, “values/quantile/0.8”: 3}, … {“values/quantile/0.2”: -3, “values/quantile/0.4”: -1, … “values/quantile/0.6”: 1, “values/quantile/0.8”: 3}, … {“values/quantile/0.2”: None, “values/quantile/0.4”: None, … “values/quantile/0.6”: None, “values/quantile/0.8”: None}, … {“values/quantile/0.2”: -3, “values/quantile/0.4”: -1, … “values/quantile/0.6”: 1, “values/quantile/0.8”: 3}, … {“values/quantile/0.2”: -3, “values/quantile/0.4”: -1, … “values/quantile/0.6”: 1, “values/quantile/0.8”: 3}, … {“values/quantile/0.2”: None, “values/quantile/0.4”: None, … “values/quantile/0.6”: None, “values/quantile/0.8”: None}, … ], … }, … schema = { … “code”: pl.Utf8, … “values/quantiles”: pl.Struct([ … pl.Field(“values/quantile/0.2”, pl.Float64), … pl.Field(“values/quantile/0.4”, pl.Float64), … pl.Field(“values/quantile/0.6”, pl.Float64), … pl.Field(“values/quantile/0.8”, pl.Float64), … ]), # pl.List(pl.Float64), … }, … ) quantile_code_metadata = generate_quantile_code_metadata(code_metadata) quantile_code_metadata.sort(“code”) shape: (26, 2) ┌───────────────────────┬──────────────┐ │ values/quantiles ┆ code │ │ — ┆ — │ │ struct[4] ┆ str │ ╞═══════════════════════╪══════════════╡ │ {null,null,null,null} ┆ dx//B │ │ {null,null,null,null} ┆ dx//D │ │ {-3.0,-1.0,1.0,3.0} ┆ dx//E │ │ {-3.0,-1.0,1.0,3.0} ┆ dx//E//_Q_1 │ │ {-3.0,-1.0,1.0,3.0} ┆ dx//E//_Q_2 │ │ … ┆ … │ │ {-3.0,-1.0,1.0,3.0} ┆ lab//F//_Q_1 │ │ {-3.0,-1.0,1.0,3.0} ┆ lab//F//_Q_2 │ │ {-3.0,-1.0,1.0,3.0} ┆ lab//F//_Q_3 │ │ {-3.0,-1.0,1.0,3.0} ┆ lab//F//_Q_4 │ │ {-3.0,-1.0,1.0,3.0} ┆ lab//F//_Q_5 │ └───────────────────────┴──────────────┘

Source code in meds_torch/utils/quantile_binning.py 
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
def generate_quantile_code_metadata(code_metadata: pl.DataFrame) -> pl.DataFrame:
    """Modifies the code_metadata DataFrame to include quantile codes.

    Args:
        code_metadata (pl.DataFrame): Current code metadata DataFrame

    Returns:
        pl.DataFrame: dataframe where we duplicate rows for each quantile code.

    Examples:
    >>> from datetime import datetime
    >>> code_metadata = pl.DataFrame(
    ...     {
    ...         "code": ["lab//A", "lab//C", "dx//B", "dx//E", "lab//F", "dx//D"],
    ...         "values/quantiles": [ # [[-3,-1,1,3], [-3,-1,1,3], [], [-3,-1,1,3], [-3,-1,1,3], [-3,-1,1,3]],
    ...             {"values/quantile/0.2": -3, "values/quantile/0.4": -1,
    ...                 "values/quantile/0.6": 1, "values/quantile/0.8": 3},
    ...             {"values/quantile/0.2": -3, "values/quantile/0.4": -1,
    ...                 "values/quantile/0.6": 1, "values/quantile/0.8": 3},
    ...             {"values/quantile/0.2": None, "values/quantile/0.4": None,
    ...                 "values/quantile/0.6": None, "values/quantile/0.8": None},
    ...             {"values/quantile/0.2": -3, "values/quantile/0.4": -1,
    ...                 "values/quantile/0.6": 1, "values/quantile/0.8": 3},
    ...             {"values/quantile/0.2": -3, "values/quantile/0.4": -1,
    ...                 "values/quantile/0.6": 1, "values/quantile/0.8": 3},
    ...             {"values/quantile/0.2": None, "values/quantile/0.4": None,
    ...                 "values/quantile/0.6": None, "values/quantile/0.8": None},
    ...         ],
    ...     },
    ...     schema = {
    ...         "code": pl.Utf8,
    ...         "values/quantiles": pl.Struct([
    ...             pl.Field("values/quantile/0.2", pl.Float64),
    ...             pl.Field("values/quantile/0.4", pl.Float64),
    ...             pl.Field("values/quantile/0.6", pl.Float64),
    ...             pl.Field("values/quantile/0.8", pl.Float64),
    ...         ]), # pl.List(pl.Float64),
    ...     },
    ... )
    >>> quantile_code_metadata = generate_quantile_code_metadata(code_metadata)
    >>> quantile_code_metadata.sort("code")
    shape: (26, 2)
    ┌───────────────────────┬──────────────┐
    │ values/quantiles      ┆ code         │
    │ ---                   ┆ ---          │
    │ struct[4]             ┆ str          │
    ╞═══════════════════════╪══════════════╡
    │ {null,null,null,null} ┆ dx//B        │
    │ {null,null,null,null} ┆ dx//D        │
    │ {-3.0,-1.0,1.0,3.0}   ┆ dx//E        │
    │ {-3.0,-1.0,1.0,3.0}   ┆ dx//E//_Q_1  │
    │ {-3.0,-1.0,1.0,3.0}   ┆ dx//E//_Q_2  │
    │ …                     ┆ …            │
    │ {-3.0,-1.0,1.0,3.0}   ┆ lab//F//_Q_1 │
    │ {-3.0,-1.0,1.0,3.0}   ┆ lab//F//_Q_2 │
    │ {-3.0,-1.0,1.0,3.0}   ┆ lab//F//_Q_3 │
    │ {-3.0,-1.0,1.0,3.0}   ┆ lab//F//_Q_4 │
    │ {-3.0,-1.0,1.0,3.0}   ┆ lab//F//_Q_5 │
    └───────────────────────┴──────────────┘
    """
    # Step 1: Get the number of quantiles
    quantile_fields = code_metadata.schema["values/quantiles"].fields
    num_bins = len(quantile_fields) + 1  # Add 1 for n+1 bins
    codes_per_entry = num_bins + 1  # Add 1 for the original code

    # Get first quantile name
    first_quantile = quantile_fields[0].name
    is_null_quantile = code_metadata.select(pl.col("values/quantiles").struct.field(first_quantile).is_null())

    # Determine number of bins per row
    codes_per_entry_expr = pl.when(is_null_quantile).then(1).otherwise(codes_per_entry)

    # If custom quantiles exist, update the expression
    if "custom_quantiles" in code_metadata.columns and isinstance(
        code_metadata.schema["custom_quantiles"], pl.Struct
    ):
        custom_bins = len(code_metadata.schema["custom_quantiles"].fields) + 2
        codes_per_entry_expr = (
            pl.when(pl.col("custom_quantiles").is_not_null())
            .then(custom_bins + 2)
            .otherwise(codes_per_entry_expr)
        )

    code_metadata = code_metadata.with_columns(codes_per_entry_expr.alias("codes_per_entry_expr"))

    # Step 2: Generate rows for each code and its quantiles
    code_metadata = code_metadata.with_row_index("code_index")
    expanded_metadata = (
        code_metadata.select(
            pl.col("code").repeat_by(pl.col("codes_per_entry_expr")),
            pl.col("code_index").repeat_by(pl.col("codes_per_entry_expr")).alias("base_index"),
            pl.exclude("code", "code_index"),
        )
        .explode(["code", "base_index"])
        .with_row_index()
    )

    # Step 3: Generate binned code vocab indices and adjust vocab indices
    offset = (
        expanded_metadata.group_by("code", maintain_order=True)
        .agg(pl.col("base_index").first(), pl.col("codes_per_entry_expr").first())
        .select(
            pl.col("code"),
            pl.col("codes_per_entry_expr").cum_sum().alias("offset") - pl.col("codes_per_entry_expr"),
        )
    )
    expanded_metadata = expanded_metadata.join(offset, on="code", how="left")
    assert expanded_metadata["base_index"].is_sorted()
    expanded_metadata = expanded_metadata.with_columns((pl.col("index") - pl.col("offset")).alias("quantile"))
    expanded_metadata = expanded_metadata.rename({"index": "code_index"}).drop(
        "codes_per_entry_expr", "offset", "code_index"
    )
    if "custom_quantiles" in expanded_metadata.columns:
        expanded_metadata = expanded_metadata.drop("custom_quantiles")

    # Step 4: Generate quantile codes
    expanded_metadata = expanded_metadata.with_columns(
        binned_code=pl.when(pl.col("quantile") != 0)
        .then(pl.concat_str(pl.col("code"), pl.lit("//_Q_"), pl.col("quantile").cast(pl.Utf8)))
        .otherwise(pl.col("code"))
    )

    # Step 5: Select and rename final columns
    final_metadata = expanded_metadata.drop("base_index", "code", "quantile").rename({"binned_code": "code"})

    return final_metadata

main(cfg)

Bins the numeric values and collapses the bin number into the code name.

DO NOT RUN THIS WITH PARALLELISM as it will recursively perform quantile binning N workers times.

Source code in meds_torch/utils/quantile_binning.py 
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
@hydra.main(
    version_base=None, config_path=str(PREPROCESS_CONFIG_YAML.parent), config_name=PREPROCESS_CONFIG_YAML.stem
)
def main(cfg: DictConfig):
    """Bins the numeric values and collapses the bin number into the code name.

    DO NOT RUN THIS WITH PARALLELISM as it will recursively perform quantile binning N workers times.
    """

    def normalize(df, code_metadata, code_modifiers=None):
        return quantile_normalize(
            df,
            code_metadata,
            code_modifiers=code_modifiers,
            custom_quantiles=cfg.stage_cfg.get("custom_quantiles", {}),
        )

    map_over(cfg, compute_fn=normalize)

    custom_quantiles = cfg.stage_cfg.get("custom_quantiles", {})

    metadata_input_dir = Path(cfg.stage_cfg.metadata_input_dir)
    code_metadata = pl.read_parquet(metadata_input_dir / "codes.parquet", use_pyarrow=True)
    quantile_code_metadata = convert_metadata_codes_to_discrete_quantiles(code_metadata, custom_quantiles)

    output_fp = metadata_input_dir / "codes.parquet"
    output_fp.parent.mkdir(parents=True, exist_ok=True)
    logger.info(f"Indices assigned. Writing to {output_fp}")
    quantile_code_metadata.write_parquet(output_fp, use_pyarrow=True)

process_quantiles(df)

Process quantiles in a DataFrame, using custom quantiles if available.

Parameters:

Name Type Description Default
df DataFrame

A Polars DataFrame containing columns for values/quantiles and optionally custom_quantiles

 required

Returns:

Type Description
DataFrame

A DataFrame with processed quantiles and updated code column

Examples:

import polars as pl df = pl.DataFrame({ … “code”: [“lab//A”, “lab//B”, “lab//C”, “lab//D”], … “numeric_value”: [-1.0, 2.0, None, 0.0], … “values/quantiles”: [ … {“values/quantile/0.2”: 0, “values/quantile/0.4”: 1, … “values/quantile/0.6”: 2, “values/quantile/0.8”: 3}, … {“values/quantile/0.2”: 0, “values/quantile/0.4”: 1, … “values/quantile/0.6”: 2, “values/quantile/0.8”: 3}, … {“values/quantile/0.2”: 0, “values/quantile/0.4”: 1, … “values/quantile/0.6”: 2, “values/quantile/0.8”: 3}, … {“values/quantile/0.2”: 0, “values/quantile/0.4”: 1, … “values/quantile/0.6”: 2, “values/quantile/0.8”: 3} … ], … “custom_quantiles”: [None, … {“values/quantile/0.5”: 1.5}, … None, … None … ], … “code/vocab_index”: [0, 1, 2, 3], … }) result = process_quantiles(df) result.select([“code”, “numeric_value”]) shape: (4, 2) ┌──────────────┬───────────────┐ │ code ┆ numeric_value │ │ — ┆ — │ │ str ┆ f64 │ ╞══════════════╪═══════════════╡ │ lab//A//_Q_1 ┆ -1.0 │ │ lab//B//_Q_2 ┆ 2.0 │ │ lab//C ┆ null │ │ lab//D//_Q_1 ┆ 0.0 │ └──────────────┴───────────────┘

Test with only custom quantiles

df_custom = pl.DataFrame({ … “code”: [“lab//A”, “lab//A”, “lab//A”], … “numeric_value”: [-0.5, 1.0, 4.0], … “values/quantiles”: [ … {“values/quantile/0.2”: 0, “values/quantile/0.4”: 1, … “values/quantile/0.6”: 2, “values/quantile/0.8”: 3}, … {“values/quantile/0.2”: 0, “values/quantile/0.4”: 1, … “values/quantile/0.6”: 2, “values/quantile/0.8”: 3}, … {“values/quantile/0.2”: 0, “values/quantile/0.4”: 1, … “values/quantile/0.6”: 2, “values/quantile/0.8”: 3}, … ], … “custom_quantiles”: [ … None, … None, … None, … ], … “code/vocab_index”: [0, 0, 0], … }) result_custom = process_quantiles(df_custom) result_custom.select([“code”, “numeric_value”]) shape: (3, 2) ┌──────────────┬───────────────┐ │ code ┆ numeric_value │ │ — ┆ — │ │ str ┆ f64 │ ╞══════════════╪═══════════════╡ │ lab//A//_Q_1 ┆ -0.5 │ │ lab//A//_Q_2 ┆ 1.0 │ │ lab//A//_Q_5 ┆ 4.0 │ └──────────────┴───────────────┘

Source code in meds_torch/utils/quantile_binning.py 
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
def process_quantiles(df: pl.DataFrame) -> pl.DataFrame:
    """Process quantiles in a DataFrame, using custom quantiles if available.

    Args:
        df: A Polars DataFrame containing columns for values/quantiles and optionally custom_quantiles

    Returns:
        A DataFrame with processed quantiles and updated code column

    Examples:
    >>> import polars as pl
    >>> df = pl.DataFrame({
    ...     "code": ["lab//A", "lab//B", "lab//C", "lab//D"],
    ...     "numeric_value": [-1.0, 2.0, None, 0.0],
    ...     "values/quantiles": [
    ...         {"values/quantile/0.2": 0, "values/quantile/0.4": 1,
    ...          "values/quantile/0.6": 2, "values/quantile/0.8": 3},
    ...         {"values/quantile/0.2": 0, "values/quantile/0.4": 1,
    ...          "values/quantile/0.6": 2, "values/quantile/0.8": 3},
    ...         {"values/quantile/0.2": 0, "values/quantile/0.4": 1,
    ...          "values/quantile/0.6": 2, "values/quantile/0.8": 3},
    ...         {"values/quantile/0.2": 0, "values/quantile/0.4": 1,
    ...          "values/quantile/0.6": 2, "values/quantile/0.8": 3}
    ...     ],
    ...     "custom_quantiles": [None,
    ...         {"values/quantile/0.5": 1.5},
    ...         None,
    ...         None
    ...     ],
    ...     "code/vocab_index": [0, 1, 2, 3],
    ... })
    >>> result = process_quantiles(df)
    >>> result.select(["code", "numeric_value"])
    shape: (4, 2)
    ┌──────────────┬───────────────┐
    │ code         ┆ numeric_value │
    │ ---          ┆ ---           │
    │ str          ┆ f64           │
    ╞══════════════╪═══════════════╡
    │ lab//A//_Q_1 ┆ -1.0          │
    │ lab//B//_Q_2 ┆ 2.0           │
    │ lab//C       ┆ null          │
    │ lab//D//_Q_1 ┆ 0.0           │
    └──────────────┴───────────────┘
    >>> # Test with only custom quantiles
    >>> df_custom = pl.DataFrame({
    ...     "code": ["lab//A", "lab//A", "lab//A"],
    ...     "numeric_value": [-0.5, 1.0, 4.0],
    ...     "values/quantiles": [
    ...         {"values/quantile/0.2": 0, "values/quantile/0.4": 1,
    ...          "values/quantile/0.6": 2, "values/quantile/0.8": 3},
    ...         {"values/quantile/0.2": 0, "values/quantile/0.4": 1,
    ...          "values/quantile/0.6": 2, "values/quantile/0.8": 3},
    ...         {"values/quantile/0.2": 0, "values/quantile/0.4": 1,
    ...          "values/quantile/0.6": 2, "values/quantile/0.8": 3},
    ...     ],
    ...     "custom_quantiles": [
    ...         None,
    ...         None,
    ...         None,
    ...     ],
    ...     "code/vocab_index": [0, 0, 0],
    ... })
    >>> result_custom = process_quantiles(df_custom)
    >>> result_custom.select(["code", "numeric_value"])
    shape: (3, 2)
    ┌──────────────┬───────────────┐
    │ code         ┆ numeric_value │
    │ ---          ┆ ---           │
    │ str          ┆ f64           │
    ╞══════════════╪═══════════════╡
    │ lab//A//_Q_1 ┆ -0.5          │
    │ lab//A//_Q_2 ┆ 1.0           │
    │ lab//A//_Q_5 ┆ 4.0           │
    └──────────────┴───────────────┘
    """
    # Use custom_quantiles if available, otherwise use values/quantiles
    df = df.with_columns(
        effective_quantiles=pl.when(pl.col("custom_quantiles").is_not_null())
        .then(pl.col("custom_quantiles"))
        .otherwise(pl.col("values/quantiles"))
    )

    # Unnest the effective_quantiles and calculate the quantile
    quantile_columns = pl.selectors.starts_with("values/quantile/")
    df = df.unnest("effective_quantiles").with_columns(
        quantile=pl.when(pl.col("numeric_value").is_not_null()).then(
            pl.sum_horizontal(quantile_columns.lt(pl.col("numeric_value"))).add(1)
        )
    )

    # Create the new code with quantile information
    code_quantile_concat = pl.concat_str(pl.col("code"), pl.lit("//_Q_"), pl.col("quantile"))
    df = df.with_columns(
        code=pl.when(pl.col("quantile").is_not_null()).then(code_quantile_concat).otherwise(pl.col("code"))
    )

    # Clean up intermediate columns
    df = df.drop("quantile", "values/quantiles", "custom_quantiles", quantile_columns)

    return df

quantile_normalize(df, code_metadata, code_modifiers=None, custom_quantiles={})

Normalize a MEDS dataset across both categorical and continuous dimensions.

This function expects a MEDS dataset in flattened form, with columns for: - subject_id - time - code - numeric_value

In addition, the code_metadata dataset should contain information about the codes in the MEDS dataset, including the mandatory columns: - code (categorical) - Any code_modifiers columns, if specified - values/quantiles

Parameters:

Name Type Description Default
df LazyFrame

The MEDS dataset to normalize. See above for the expected schema.

 required
code_metadata DataFrame

Metadata about the codes in the MEDS dataset. See above for the expected schema.

 required
code_modifiers list[str] | None

Additional columns to join on, which will be discarded from the output dataframe.

 None

Returns:

Type Description
LazyFrame

The normalized MEDS dataset, with the schema described above.

Examples:

>>> from datetime import datetime
>>> MEDS_df = pl.DataFrame(
...     {
...         "subject_id": [1, 1, 1, 2, 2, 2, 3],
...         "time": [
...             datetime(2021, 1, 1),
...             datetime(2021, 1, 1),
...             datetime(2021, 1, 2),
...             datetime(2022, 10, 2),
...             datetime(2022, 10, 2),
...             datetime(2022, 10, 2),
...             datetime(2022, 10, 2),
...         ],
...         "code": ["lab//A", "lab//A", "dx//B", "lab//A", "dx//D", "lab//C", "lab//F"],
...         "numeric_value": [1, 3, None, 3, None, None, None],
...     },
...     schema = {
...         "subject_id": pl.UInt32,
...         "time": pl.Datetime,
...         "code": pl.Utf8,
...         "numeric_value": pl.Float64,
...    },
... )
>>> code_metadata = pl.DataFrame(
...     {
...         "code": ["lab//A", "lab//C", "dx//B", "dx//E", "lab//F"],
...         "values/quantiles": [ # [[-3,-1,1,3], [-3,-1,1,3], [], [-3,-1,1,3], [-3,-1,1,3]],
...             {"values/quantile/0.2": -3, "values/quantile/0.4": -1,
...                 "values/quantile/0.6": 1, "values/quantile/0.8": 3},
...             {"values/quantile/0.2": -3, "values/quantile/0.4": -1,
...                 "values/quantile/0.6": 1, "values/quantile/0.8": 3},
...             {"values/quantile/0.2": None, "values/quantile/0.4": None,
...                 "values/quantile/0.6": None, "values/quantile/0.8": None},
...             {"values/quantile/0.2": -3, "values/quantile/0.4": -1,
...                 "values/quantile/0.6": 1, "values/quantile/0.8": 3},
...             {"values/quantile/0.2": -3, "values/quantile/0.4": -1,
...                 "values/quantile/0.6": 1, "values/quantile/0.8": 3},
...         ],
...     },
...     schema = {
...         "code": pl.Utf8,
...         "values/quantiles": pl.Struct([
...             pl.Field("values/quantile/0.2", pl.Float64),
...             pl.Field("values/quantile/0.4", pl.Float64),
...             pl.Field("values/quantile/0.6", pl.Float64),
...             pl.Field("values/quantile/0.8", pl.Float64),
...         ]), # pl.List(pl.Float64),
...     },
... )
>>> code_metadata
shape: (5, 2)
┌────────┬───────────────────────┐
│ code   ┆ values/quantiles      │
│ ---    ┆ ---                   │
│ str    ┆ struct[4]             │
╞════════╪═══════════════════════╡
│ lab//A ┆ {-3.0,-1.0,1.0,3.0}   │
│ lab//C ┆ {-3.0,-1.0,1.0,3.0}   │
│ dx//B  ┆ {null,null,null,null} │
│ dx//E  ┆ {-3.0,-1.0,1.0,3.0}   │
│ lab//F ┆ {-3.0,-1.0,1.0,3.0}   │
└────────┴───────────────────────┘
>>> quantile_normalize(MEDS_df.lazy(), code_metadata).collect().sort("subject_id", "time", "code")
shape: (7, 4)
┌────────────┬─────────────────────┬──────────────┬───────────────┐
│ subject_id ┆ time                ┆ code         ┆ numeric_value │
│ ---        ┆ ---                 ┆ ---          ┆ ---           │
│ u32        ┆ datetime[μs]        ┆ str          ┆ f64           │
╞════════════╪═════════════════════╪══════════════╪═══════════════╡
│ 1          ┆ 2021-01-01 00:00:00 ┆ lab//A//_Q_3 ┆ 1.0           │
│ 1          ┆ 2021-01-01 00:00:00 ┆ lab//A//_Q_4 ┆ 3.0           │
│ 1          ┆ 2021-01-02 00:00:00 ┆ dx//B        ┆ null          │
│ 2          ┆ 2022-10-02 00:00:00 ┆ dx//D        ┆ null          │
│ 2          ┆ 2022-10-02 00:00:00 ┆ lab//A//_Q_4 ┆ 3.0           │
│ 2          ┆ 2022-10-02 00:00:00 ┆ lab//C       ┆ null          │
│ 3          ┆ 2022-10-02 00:00:00 ┆ lab//F       ┆ null          │
└────────────┴─────────────────────┴──────────────┴───────────────┘
>>> custom_quantiles = {"lab//A": {"values/quantile/0.5": 2}}
>>> quantile_normalize(MEDS_df.lazy(), code_metadata, custom_quantiles=custom_quantiles
...     ).collect().sort("subject_id", "time", "code")
shape: (7, 4)
┌────────────┬─────────────────────┬──────────────┬───────────────┐
│ subject_id ┆ time                ┆ code         ┆ numeric_value │
│ ---        ┆ ---                 ┆ ---          ┆ ---           │
│ u32        ┆ datetime[μs]        ┆ str          ┆ f64           │
╞════════════╪═════════════════════╪══════════════╪═══════════════╡
│ 1          ┆ 2021-01-01 00:00:00 ┆ lab//A//_Q_1 ┆ 1.0           │
│ 1          ┆ 2021-01-01 00:00:00 ┆ lab//A//_Q_2 ┆ 3.0           │
│ 1          ┆ 2021-01-02 00:00:00 ┆ dx//B        ┆ null          │
│ 2          ┆ 2022-10-02 00:00:00 ┆ dx//D        ┆ null          │
│ 2          ┆ 2022-10-02 00:00:00 ┆ lab//A//_Q_2 ┆ 3.0           │
│ 2          ┆ 2022-10-02 00:00:00 ┆ lab//C       ┆ null          │
│ 3          ┆ 2022-10-02 00:00:00 ┆ lab//F       ┆ null          │
└────────────┴─────────────────────┴──────────────┴───────────────┘
Source code in meds_torch/utils/quantile_binning.py 
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
def quantile_normalize(
    df: pl.LazyFrame,
    code_metadata: pl.DataFrame,
    code_modifiers: list[str] | None = None,
    custom_quantiles: dict = {},
) -> pl.LazyFrame:
    """Normalize a MEDS dataset across both categorical and continuous dimensions.

    This function expects a MEDS dataset in flattened form, with columns for:
      - `subject_id`
      - `time`
      - `code`
      - `numeric_value`

    In addition, the `code_metadata` dataset should contain information about the codes in the MEDS dataset,
    including the mandatory columns:
      - `code` (`categorical`)
      - Any `code_modifiers` columns, if specified
      - `values/quantiles`

    Args:
        df: The MEDS dataset to normalize. See above for the expected schema.
        code_metadata: Metadata about the codes in the MEDS dataset. See above for the expected schema.
        code_modifiers: Additional columns to join on, which will be discarded from the output dataframe.

    Returns:
        The normalized MEDS dataset, with the schema described above.

    Examples:
        >>> from datetime import datetime
        >>> MEDS_df = pl.DataFrame(
        ...     {
        ...         "subject_id": [1, 1, 1, 2, 2, 2, 3],
        ...         "time": [
        ...             datetime(2021, 1, 1),
        ...             datetime(2021, 1, 1),
        ...             datetime(2021, 1, 2),
        ...             datetime(2022, 10, 2),
        ...             datetime(2022, 10, 2),
        ...             datetime(2022, 10, 2),
        ...             datetime(2022, 10, 2),
        ...         ],
        ...         "code": ["lab//A", "lab//A", "dx//B", "lab//A", "dx//D", "lab//C", "lab//F"],
        ...         "numeric_value": [1, 3, None, 3, None, None, None],
        ...     },
        ...     schema = {
        ...         "subject_id": pl.UInt32,
        ...         "time": pl.Datetime,
        ...         "code": pl.Utf8,
        ...         "numeric_value": pl.Float64,
        ...    },
        ... )
        >>> code_metadata = pl.DataFrame(
        ...     {
        ...         "code": ["lab//A", "lab//C", "dx//B", "dx//E", "lab//F"],
        ...         "values/quantiles": [ # [[-3,-1,1,3], [-3,-1,1,3], [], [-3,-1,1,3], [-3,-1,1,3]],
        ...             {"values/quantile/0.2": -3, "values/quantile/0.4": -1,
        ...                 "values/quantile/0.6": 1, "values/quantile/0.8": 3},
        ...             {"values/quantile/0.2": -3, "values/quantile/0.4": -1,
        ...                 "values/quantile/0.6": 1, "values/quantile/0.8": 3},
        ...             {"values/quantile/0.2": None, "values/quantile/0.4": None,
        ...                 "values/quantile/0.6": None, "values/quantile/0.8": None},
        ...             {"values/quantile/0.2": -3, "values/quantile/0.4": -1,
        ...                 "values/quantile/0.6": 1, "values/quantile/0.8": 3},
        ...             {"values/quantile/0.2": -3, "values/quantile/0.4": -1,
        ...                 "values/quantile/0.6": 1, "values/quantile/0.8": 3},
        ...         ],
        ...     },
        ...     schema = {
        ...         "code": pl.Utf8,
        ...         "values/quantiles": pl.Struct([
        ...             pl.Field("values/quantile/0.2", pl.Float64),
        ...             pl.Field("values/quantile/0.4", pl.Float64),
        ...             pl.Field("values/quantile/0.6", pl.Float64),
        ...             pl.Field("values/quantile/0.8", pl.Float64),
        ...         ]), # pl.List(pl.Float64),
        ...     },
        ... )
        >>> code_metadata
        shape: (5, 2)
        ┌────────┬───────────────────────┐
        │ code   ┆ values/quantiles      │
        │ ---    ┆ ---                   │
        │ str    ┆ struct[4]             │
        ╞════════╪═══════════════════════╡
        │ lab//A ┆ {-3.0,-1.0,1.0,3.0}   │
        │ lab//C ┆ {-3.0,-1.0,1.0,3.0}   │
        │ dx//B  ┆ {null,null,null,null} │
        │ dx//E  ┆ {-3.0,-1.0,1.0,3.0}   │
        │ lab//F ┆ {-3.0,-1.0,1.0,3.0}   │
        └────────┴───────────────────────┘
        >>> quantile_normalize(MEDS_df.lazy(), code_metadata).collect().sort("subject_id", "time", "code")
        shape: (7, 4)
        ┌────────────┬─────────────────────┬──────────────┬───────────────┐
        │ subject_id ┆ time                ┆ code         ┆ numeric_value │
        │ ---        ┆ ---                 ┆ ---          ┆ ---           │
        │ u32        ┆ datetime[μs]        ┆ str          ┆ f64           │
        ╞════════════╪═════════════════════╪══════════════╪═══════════════╡
        │ 1          ┆ 2021-01-01 00:00:00 ┆ lab//A//_Q_3 ┆ 1.0           │
        │ 1          ┆ 2021-01-01 00:00:00 ┆ lab//A//_Q_4 ┆ 3.0           │
        │ 1          ┆ 2021-01-02 00:00:00 ┆ dx//B        ┆ null          │
        │ 2          ┆ 2022-10-02 00:00:00 ┆ dx//D        ┆ null          │
        │ 2          ┆ 2022-10-02 00:00:00 ┆ lab//A//_Q_4 ┆ 3.0           │
        │ 2          ┆ 2022-10-02 00:00:00 ┆ lab//C       ┆ null          │
        │ 3          ┆ 2022-10-02 00:00:00 ┆ lab//F       ┆ null          │
        └────────────┴─────────────────────┴──────────────┴───────────────┘
        >>> custom_quantiles = {"lab//A": {"values/quantile/0.5": 2}}
        >>> quantile_normalize(MEDS_df.lazy(), code_metadata, custom_quantiles=custom_quantiles
        ...     ).collect().sort("subject_id", "time", "code")
        shape: (7, 4)
        ┌────────────┬─────────────────────┬──────────────┬───────────────┐
        │ subject_id ┆ time                ┆ code         ┆ numeric_value │
        │ ---        ┆ ---                 ┆ ---          ┆ ---           │
        │ u32        ┆ datetime[μs]        ┆ str          ┆ f64           │
        ╞════════════╪═════════════════════╪══════════════╪═══════════════╡
        │ 1          ┆ 2021-01-01 00:00:00 ┆ lab//A//_Q_1 ┆ 1.0           │
        │ 1          ┆ 2021-01-01 00:00:00 ┆ lab//A//_Q_2 ┆ 3.0           │
        │ 1          ┆ 2021-01-02 00:00:00 ┆ dx//B        ┆ null          │
        │ 2          ┆ 2022-10-02 00:00:00 ┆ dx//D        ┆ null          │
        │ 2          ┆ 2022-10-02 00:00:00 ┆ lab//A//_Q_2 ┆ 3.0           │
        │ 2          ┆ 2022-10-02 00:00:00 ┆ lab//C       ┆ null          │
        │ 3          ┆ 2022-10-02 00:00:00 ┆ lab//F       ┆ null          │
        └────────────┴─────────────────────┴──────────────┴───────────────┘
    """
    # TODO: add support for original values/mean and values/std normalization of continuous values
    df = convert_to_discrete_quantiles(df.collect(), code_metadata, custom_quantiles)
    return df.lazy()