AttributeError when using threshold_selection_tool_from_labels_table with DuckDBLinker

[Ahosseinzadeh723]

I'm currently working with splink 3.9.13 and trying to utilize the threshold_selection_tool_from_labels_table function as demonstrated in the documentation found here

I've prepared my labeled table as instructed here

However, when I attempt to execute the following line of code:

linker.threshold_selection_tool_from_labels_table(labels_table, add_metrics=['f1'])

I receive the following error:

'DuckDBLinker' object has no attribute 'threshold_selection_tool_from_labels_table'

AttributeError: 'DuckDBLinker' object has no attribute 'threshold_selection_tool_from_labels_table'

This suggests that the DuckDBLinker object does not have the threshold_selection_tool_from_labels_table method. Is this functionality not supported for DuckDBLinker, or is this an issue with the library version or documentation?

Thanks!

[RobinL]

Edit 2: The problem is that this tool appeared in 3.9.14, so you need to upgrade! pip install --upgrade splink

The following is a script that I tested as working on 3.9.14

import splink.duckdb.comparison_library as cl
from splink.datasets import splink_dataset_labels, splink_datasets
from splink.duckdb.blocking_rule_library import block_on
from splink.duckdb.linker import DuckDBLinker

df = splink_datasets.fake_1000
settings = {
    "link_type": "dedupe_only",
    "blocking_rules_to_generate_predictions": [
        block_on(["first_name"]),
        block_on(["surname"]),
    ],
    "comparisons": [
        cl.jaro_winkler_at_thresholds(
            "first_name", 0.9, term_frequency_adjustments=True
        ),
        cl.exact_match("surname"),
        cl.levenshtein_at_thresholds("dob"),
        cl.exact_match("city", term_frequency_adjustments=True),
        cl.exact_match("email"),
    ],
}

linker = DuckDBLinker(df, settings)

linker.estimate_probability_two_random_records_match(
    [block_on(["first_name", "surname", "dob"])], recall=0.7
)

linker.estimate_parameters_using_expectation_maximisation(block_on(["first_name"]))

linker.estimate_parameters_using_expectation_maximisation(block_on(["surname"]))

labels_splink_df = linker.register_labels_table(splink_dataset_labels.fake_1000_labels)
linker.threshold_selection_tool_from_labels_table(labels_splink_df)