Usage
outlierdetection(
data,
vars,
method_category = "composite",
univariate_methods = "zscore_robust",
multivariate_methods = "mahalanobis",
composite_threshold = 0.5,
zscore_threshold = 3.29,
iqr_multiplier = 1.7,
confidence_level = 0.999,
show_outlier_table = TRUE,
show_method_comparison = TRUE,
show_exclusion_summary = TRUE,
show_visualization = TRUE,
show_interpretation = TRUE
)Arguments
- data
The data as a data frame.
- vars
Continuous variables to analyze for outliers. The module will detect outliers based on the selected variables using the chosen detection methods.
- method_category
Category of outlier detection methods to use. Univariate methods analyze each variable separately, multivariate methods consider relationships between variables, and composite combines multiple methods for robust detection.
- univariate_methods
Specific univariate method for outlier detection when univariate category is selected.
- multivariate_methods
Specific multivariate method for outlier detection when multivariate category is selected.
- composite_threshold
Threshold for composite outlier score (0.1-1.0). Default 0.5 means observations classified as outliers by at least half of the methods are considered outliers.
- zscore_threshold
Threshold for Z-score based methods. Default 3.29 corresponds to 99.9\ observations).
- iqr_multiplier
Multiplier for IQR-based outlier detection. Default 1.7 is more conservative than Tukey's 1.5, reducing false positive detection.
- confidence_level
Confidence level for interval-based methods (ETI, HDI). Default 99.9\
show_outlier_tableDisplay a comprehensive table of outlier detection results including outlier scores, distances, and classification for each observation.
show_method_comparisonCompare results across different detection methods when using composite approach.
show_exclusion_summaryProvide summary of observations recommended for exclusion and impact analysis.
show_visualizationGenerate plots showing outlier detection results and distribution of outlier scores.
show_interpretationDisplay detailed interpretation of outlier detection results and methodological notes.
A results object containing:
results$todo | a html | ||||
results$plot | an image | ||||
results$outlier_table | a html | ||||
results$method_comparison | a html | ||||
results$exclusion_summary | a html | ||||
results$interpretation | a html |