Skip to contents

Check outliers in a datras_raw object

Source: R/outliers.R
check_outliers.Rd

Rule-based quality control for key variables in a datras_raw object. Optionally adds percentile-based extreme-value flagging (group-wise).

Usage

check_outliers(
  x,
  vars = NULL,
  strict = TRUE,
  pct = FALSE,
  pct_probs = c(0.01, 0.99),
  pct_by = list(HH = c("Survey", "Quarter", "Gear", "Ship"), HL = c("Survey", "Quarter",
    "Gear", "Valid_Aphia"), CA = c("Survey", "Quarter", "Gear", "Valid_Aphia")),
  pct_vars = list(HH = c("HaulDur", "Depth", "DoorSpread", "WingSpread"), HL =
    c("LngtCm"), CA = c("Age", "IndWgt", "LngtClas")),
  pct_min_n = 50,
  pct_log_vars = list(HH = character(0), HL = character(0), CA = c("IndWgt")),
  remove_extremes = FALSE,
  action = c("report", "remove"),
  verbose = TRUE
)

Arguments

x

A datras_raw object, i.e. a list with components HH, HL, and CA.

vars

Optional character vector of variable names to check. If NULL, all default rules are used. If provided, percentile checks are also limited to these variables.

strict

Logical; if TRUE, use stricter upper bounds for rule-based checks.

pct

Logical; if TRUE, also flag extreme values using percentiles.

pct_probs

Numeric length-2 vector of lower/upper probabilities, e.g. c(0.01, 0.99).

pct_by

Named list with elements HH/HL/CA giving grouping variables for percentile calculations. Only columns present in the data are used; missing columns are silently dropped rather than collapsing all groups.

pct_vars

Named list with elements HH/HL/CA giving variables to check via percentiles.

pct_min_n

Minimum number of non-missing observations required per group to compute percentiles.

pct_log_vars

Named list with elements HH/HL/CA giving variables for which percentiles are computed on log-scale.

remove_extremes

Logical; if TRUE and action = "remove", also remove hauls flagged by percentile checks. Default FALSE (safer).

action

Character; either "report" or "remove".

verbose

Logical; print a summary?

Value

A datras_raw object. The object is returned unchanged when action = "report", and with flagged hauls removed when action = "remove".

Attributes added:

  • attr(res, "outlier_report") data.frame with all flagged rows

  • attr(res, "outlier_hauls") union of all flagged haul IDs

  • attr(res, "outlier_hauls_invalid") haul IDs flagged by rule-based checks

  • attr(res, "outlier_hauls_extreme") haul IDs flagged by percentile checks