visualize\(evaluate\)leaderboard

Heatmap leaderboard across evaluation metrics

Description

Heatmap leaderboard across evaluation metrics

Usage

visualize.evaluate.leaderboard(
  results,
  metrics = NULL,
  models = NULL,
  outcomes = NULL,
  rows.by = c("models", "metrics"),
  show.values = c("none", "rank", "value"),
  digits = 2,
  cols = c("#1b7837", "#f7f7f7", "#b2182b"),
  score.cols = c("#f7f7f7", "#1b7837"),
  ...
)

Arguments

results

A PolyGeniusEvaluation object.

metrics

Optional flat filter vector. Values may be analysis names and/or metric names. Analysis names expand to all metrics present in the object for that analysis. Similarity is always excluded.

models

Optional model filter.

outcomes

Optional outcome filter.

rows.by

Whether heatmap rows represent models or metrics.

show.values

Cell text to print (“none”, “rank”, or “value”).

digits

Number of digits for value labels.

cols

Rank heatmap colors from best rank to worst rank.

score.cols

Total score annotation colors from low to high score.

…

Additional arguments forwarded to the tidyHeatmap heatmap call.

Value

A tidyHeatmap/ComplexHeatmap heatmap.

Ranking and score annotation

Cells are always colored by dense rank. Ranks are computed within each displayed analysis/metric column after applying metric direction metadata: higher-is-better metrics rank larger estimates first, lower-is-better metrics rank smaller estimates first, and target metrics rank estimates closest to their target first. Ratio metrics such as odds ratios and hazard ratios rank estimates farther from 1 first. Tied models receive the same dense rank and therefore the same score. Missing metric cells contribute zero to model score annotations. Model score annotations first average normalized rank scores within each analysis and then sum those analysis-level scores, so each analysis category contributes at most one point regardless of how many metrics it contains.

See Also

Other visualize: visualize.evaluate.calibration.curve(), visualize.evaluate.calibration.slope.intercept(), visualize.evaluate.compare.delta.forest(), visualize.evaluate.compare.delta.heatmap(), visualize.evaluate.confusion(), visualize.evaluate.discrimination.curve(), visualize.evaluate.distribution(), visualize.evaluate.forest(), visualize.evaluate.incremental.delta(), visualize.evaluate.metric.flags(), visualize.evaluate.metric.heatmap(), visualize.evaluate.pr(), visualize.evaluate.predicted.vs.observed(), visualize.evaluate.residuals(), visualize.evaluate.risk.strata.lift(), visualize.evaluate.risk.strata.profile(), visualize.evaluate.roc(), visualize.evaluate.similarity.heatmap()