Rhino Health SDK — Metrics Guide

Help configure and run any of the 40+ federated metrics in the rhino-health Python SDK (v2.1.x).

Context Loading

Before responding, read these reference files:

1. Metrics Reference — ../../context/metrics_reference.md All metric classes with parameters, import paths, categories, and the Quick Decision Guide.

2. Patterns & Gotchas — ../../context/patterns_and_gotchas.md Focus on §4 (Per-site vs Aggregated Metrics), §5 (Filtering), §6 (Group By), and §7 (Federated Joins).

Metric Decision Tree

Map the user's question to the right metric class:

User asks about...	Metric class	Category
Counts, frequencies	Count	Basic
Averages, means	Mean	Basic
Spread, variability	StandardDeviation, Variance	Basic
Totals, sums	Sum	Basic
Percentiles, medians, quartiles	Percentile, NPercentile	Quantile
Survival time, time-to-event	KaplanMeier	Survival
Hazard ratios, covariates + survival	Cox	Survival
ROC curves, AUC	RocAuc	ROC/AUC
ROC with confidence intervals	RocAucWithCI	ROC/AUC
Correlation between variables	Pearson, Spearman	Statistics
Inter-rater reliability	ICC	Statistics
Compare two group means	TTest	Statistics
Compare 3+ group means	OneWayANOVA	Statistics
Categorical association	ChiSquare	Statistics
2x2 contingency table	TwoByTwoTable	Epidemiology
Odds ratio	OddsRatio	Epidemiology
Risk ratio / relative risk	RiskRatio	Epidemiology
Risk difference	RiskDifference	Epidemiology
Incidence rates	Incidence	Epidemiology

If unsure, consult the full Quick Decision Guide in metrics_reference.md.

Execution Mode

Choose the right execution method based on scope:

Scope	Method	Signature
Single site/dataset	session.dataset.get_dataset_metric(dataset_uid, config)	One dataset UID (str)
Single site (shorthand)	dataset.get_metric(config)	Called on a Dataset object
Aggregated across sites	session.project.aggregate_dataset_metric(dataset_uids, config)	List[str] of UIDs — must be strings, not Dataset objects
Federated join (SQL-like)	session.project.joined_dataset_metric(config, query_datasets, filter_datasets)	List[str] UIDs for both params

All metrics are imported from rhino_health.lib.metrics (NOT rhino_health.metrics).

Filtering

Apply filters to narrow the data before computing the metric. Two approaches:

Inline FilterVariable (for simple per-column filters)

from rhino_health.lib.metrics import Mean, FilterType, FilterVariable

config = Mean(
    variable="Height",
    data_filters=[
        FilterVariable(
            data_column="Gender",
            filter_column="Gender",
            filter_value="Female",
            filter_type=FilterType.EQUALS,
        )
    ],
)

Range Filter (BETWEEN)

from rhino_health.lib.metrics import FilterBetweenRange, FilterType, FilterVariable

config = Mean(
    variable="Height",
    data_filters=[
        FilterVariable(
            data_column="Age",
            filter_column="Age",
            filter_value=FilterBetweenRange(min=18, max=65),
            filter_type=FilterType.BETWEEN,
        )
    ],
)

Grouping

Split results by one or more categorical columns:

config = Mean(
    variable="Height",
    group_by={"groupings": ["Gender"]},
)

Grouping and filtering can be combined on any metric.

Response Format

Structure every response as:

1. Recommended metric — which class and why it fits the user's question
2. Configuration — complete metric config object with correct import path
3. Execution call — the right method (get_dataset_metric, aggregate_dataset_metric, or joined_dataset_metric) with correct parameter types
4. Filtering/grouping — if the user specified subsets or breakdowns, add the appropriate data_filters and/or group_by
5. See also — point to a matching example from ../../context/examples/INDEX.md if one exists

Working Examples

Check ../../context/examples/INDEX.md for matching examples. Key ones for metrics:

• eda.py — per-site and aggregated metrics with filtering and grouping
• cox.py — Cox proportional hazard regression
• metrics_examples.py — TwoByTwoTable, OddsRatio, ChiSquare, TTest, ANOVA
• roc_analysis.py — ROC curves and confidence intervals
• aggregate_quantile.py — federated percentile calculations