feat: interface for comparer and builder

[LauraBoenchenLB]

Adds Interface for DataComparer and the corresponding Builder.

DataComparer will be used to evaluate differences between multiple MedRecords or subcohorts of MedRecords.

Added a Class for creating the Cohort Config. All the functions in the DataComparer need one or multiple configured Cohorts as input. All outputs are typed dictionaries to be able to access the values easier for the user in case they want to plot them. They will also be displayed as tables similar to the edge/node overview table.

The hypothesis tests will be selected based on the type and distribution of the MedRecordAttributes and may not be possible for all attributes.

Examples:

case_group = CohortEvaluator(
    medrecord=medrecord,
    name="control",
    cohort_group="patients",
    attributes={"age": "Age", "gender": "Gender"},
    concepts_groups={
        "diagnoses": "diagnoses",
        "medications": "medications",
        "procedures": "procedures",
    },
)

control_group = CohortEvaluator(
    medrecord=medrecord2,
    name="control",
    cohort_group="patients",
    attributes={"age": "age", "gender": "gender"},
    concepts_groups={
        "diagnoses": "Diagnoses",
        "medications": "Medications",
        "procedures": "Procedures",
    },
)

# evaluating single group
control_group.get_concept_counts()

CohortComparer.calculate_distance_concepts(case_group, control_group)

# compare between groups
CohortComparer.test_difference_cohort_attributes(
    cohorts_attribute=[case_group, control_group],
    significance_level=0.05
)

# Output
{
    "age": [
        {
            "test": "t_test",
            "Hypothesis": "There is no significant difference between the means of the two populations.",
            "not_reject": True,
            "p_value": 0.1,
        }
    ],
    "income": [
        {
            "test": "t_test",
            "Hypothesis": "There is no significant difference between the means of the two populations.",
            "not_reject": False,
            "p_value": 0.6,
        }
    ],
}

[MarIniOnz]

Great start to get our comparer!

I have a couple of overall observations.

• It would be nice to have some overall comparison tables, like number of nodes of each subcohort, number of edges between those nodes, number of different attributes... Something like what happened in the print_table attributes and so. I do not know if this the idea behind the "Full_report"

• Do we want to have a comparer that can take two medrecords or two subcohorts or separate them? Could be better to have separated so it is not so confusing. That is a question that can be debated in the MedModels Weekly.

[MarIniOnz]

More comments

[MarIniOnz]

Some comments and explanations needed! I really like the typing, makes everything much easier to understand!

[JabobKrauskopf]

Noice! Much better. Some stuff as discussed in person

[MarIniOnz]

Still need my previous comments to be resolved!

[MarIniOnz]

[JabobKrauskopf]

discussed in slack

[MarIniOnz]

Awesome change, I think it has a better interface this way!

A couple of comments!

[JabobKrauskopf]

Approval to get things rolling