Find outliers in trait data (IQR rule)

Flags values below \(Q_1 - k \cdot IQR\) or above \(Q_3 + k \cdot IQR\), where \(Q_1\), \(Q_3\) are the quartiles and \(k\) is multiplier (Tukey's fences; 1.5 by default, 3 for only the extreme values). The rule is applied within each group of by, usually each location, because locations have different means.

Usage

find_outliers(
  data,
  traits,
  multiplier = 1.5,
  by = if ("location" %in% names(data)) "location" else NULL,
  id_cols = c("location", "plot_name", "plots", "plot_number", "all_entries",
    "accession", "row", "col")
)

Arguments

data

A data frame with one row per plot.

traits

Names of the numeric trait columns.

multiplier

The IQR multiplier \(k\).

by

Optional grouping column(s); by default "location" when the data have it.

id_cols

Columns that identify the plot in the report (those present in data).

Value

A tibble with one row per outlier: the identifying columns, the group, trait, value, the fences lower and upper, and side ("low" or "high").

Examples

d <- data.frame(plots = 1:20, all_entries = paste0("G", 1:20),
                yield = c(rnorm(18, 50, 2), 80, 10))
find_outliers(d, "yield")
#> # A tibble: 2 × 8
#>   plots all_entries trait value lower upper side   .row
#>   <int> <chr>       <chr> <dbl> <dbl> <dbl> <chr> <int>
#> 1    19 G19         yield    80  40.5  58.4 high     19
#> 2    20 G20         yield    10  40.5  58.4 low      20