Skip to contents

Handle Missing Values

Source: R/handle_missing.R
handle_missing.Rd

Automatically chooses between imputation and row deletion based on two criteria: dataset size (n) and overall missing rate.

Usage

handle_missing(
  analysis_result = NULL,
  data = NULL,
  method = c("linear", "knn"),
  knn_k = 5L,
  verbose = TRUE
)

Arguments

analysis_result

The list returned by missing_analysis(). Alternatively, pass a plain data.frame to data.

data

A data.frame. Used only when analysis_result is NULL.

method

Character. Imputation method: "linear" (default) or "knn".

knn_k

Integer. Number of neighbours for KNN (default 5).

verbose

Logical. Print progress messages (default TRUE).

Value

An invisible list with elements:

data_clean

The cleaned data.frame.

action

Character: "drop" or "impute".

method

Imputation method used (or "none").

imputed_mask

Logical data.frame marking imputed cells (same dimensions as data_clean). All FALSE when action is "drop".

n_imputed

Integer. Number of cells that were imputed.

report

A short character string summarising the action taken.

Details

Decision rules:

  • n <= 50 -> always impute (data too small to lose rows)

  • n > 50 and missing > 5% -> impute

  • n > 50 and missing <= 5% -> drop rows with NA

When imputing, linear interpolation is the default; KNN imputation is available as an alternative.

Examples

data(airquality)
ana  <- missing_analysis(airquality, plot = FALSE)
#> 
#> ============================================================
#>   STEP 3/7 : Missing Value Analysis
#> ============================================================
#> 
#>   Rows    : 153  |  Cols: 6  |  Total cells: 918
#>   Missing : 4.79%  (44 cells)  |  2 / 6 cols affected
#>   Action  : DROP  -- missing = 4.79% <= 5% threshold, safe to remove rows
#> 
#> ------------------------------------------------------------
#>   Missing per column
#> ------------------------------------------------------------
#>  variable n_missing pct_missing
#>     Ozone        37      24.18%
#>   Solar.R         7       4.58%
#> 
clean <- handle_missing(ana)
#> 
#> ============================================================
#>   STEP 5/9 : Handle Missing Values  [DROP]
#> ============================================================
#> 
#>   Missing : 4.79%  |  Reason: 50 < n = 153 < 1000 and missing = 4.79% <= 5% -> drop rows
#> 
#>   Rows before : 153  ->  after: 111  (-42 removed)
#> 
clean$report
#> [1] "Action: ROW DELETION  |  Removed 42 rows  |  50 < n = 153 < 1000 and missing = 4.79% <= 5% -> drop rows"