QC-15: Detect statistical outliers in numeric columns

For each column whose resolved type is "numeric", applies up to two outlier detection methods (combined with logical OR):

1. Z-score: values whose absolute Z-score exceeds max_z_score are flagged.

2. IQR fence: values below Q1 - k * IQR or above Q3 + k * IQR (where k = iqr_fence_multiplier) are flagged.

Both thresholds support per-column overrides via config$column_rules. A column is skipped (PASS with a note) when neither threshold is configured or when it has fewer than four parseable values.

Usage

check_outliers(df, config, types = NULL)

Arguments

df

A data frame with all columns as character vectors (as returned by read_dataset).

config

Named list. Merged configuration as returned by load_config.

types

Optional named character vector of pre-resolved column types; see check_inferred_types.

Value

A list of dq_result objects, one per numeric column. Status is "FAIL" when outliers are detected; "PASS" otherwise. Returns an empty list if no numeric columns are found.

Examples

cfg_dir <- system.file("demonstrations/config", package = "dqcheckr")
cfg  <- load_config("starwars_csv", config_dir = cfg_dir)
path <- system.file("demonstrations/data/starwars.csv", package = "dqcheckr")
df   <- read_dataset(path, cfg)
check_outliers(df, cfg)
#> [[1]]
#> [[1]]$check_id
#> [1] "QC-15"
#> 
#> [[1]]$check_name
#> [1] "Outlier detection"
#> 
#> [[1]]$column
#> [1] "height"
#> 
#> [[1]]$status
#> [1] "PASS"
#> 
#> [[1]]$observed
#> [1] "No outlier threshold configured."
#> 
#> [[1]]$threshold
#> [1] NA
#> 
#> [[1]]$message
#> [1] "Column 'height': outlier check skipped (no threshold)."
#> 
#> 
#> [[2]]
#> [[2]]$check_id
#> [1] "QC-15"
#> 
#> [[2]]$check_name
#> [1] "Outlier detection"
#> 
#> [[2]]$column
#> [1] "mass"
#> 
#> [[2]]$status
#> [1] "PASS"
#> 
#> [[2]]$observed
#> [1] "No outlier threshold configured."
#> 
#> [[2]]$threshold
#> [1] NA
#> 
#> [[2]]$message
#> [1] "Column 'mass': outlier check skipped (no threshold)."
#> 
#> 
#> [[3]]
#> [[3]]$check_id
#> [1] "QC-15"
#> 
#> [[3]]$check_name
#> [1] "Outlier detection"
#> 
#> [[3]]$column
#> [1] "birth_year"
#> 
#> [[3]]$status
#> [1] "PASS"
#> 
#> [[3]]$observed
#> [1] "No outlier threshold configured."
#> 
#> [[3]]$threshold
#> [1] NA
#> 
#> [[3]]$message
#> [1] "Column 'birth_year': outlier check skipped (no threshold)."
#> 
#>