Skip to contents

Robust Control Limits via Alternative Methods

Source: R/spm.R
spm.limit.robust.Rd

Computes T-squared or SPE control limits using methods beyond the standard parametric approach: empirical quantile, bootstrap, or kernel density estimation. Useful when the parametric chi-squared assumption is questionable.

Usage

spm.limit.robust(
  values,
  ncomp = NULL,
  alpha = 0.05,
  type = c("t2", "spe"),
  method = c("parametric", "empirical", "bootstrap", "kde"),
  n.bootstrap = 1000,
  seed = 42
)

Arguments

values

Numeric vector of Phase I statistic values (T-squared or SPE).

ncomp

Number of components (required for T-squared; ignored for SPE).

alpha

Significance level (default 0.05).

type

Which statistic: "t2" or "spe".

method

Estimation method: "parametric", "empirical", "bootstrap", or "kde".

n.bootstrap

Number of bootstrap resamples (default 1000; only used when method = "bootstrap").

seed

Random seed for bootstrap (default 42).

Value

A list with components:

ucl

Upper control limit

alpha

Significance level used

description

Character string describing the method

method

The method used

See also

spm.phase1 for standard control limits

Examples

# \donttest{
set.seed(1)
t2_values <- rchisq(100, df = 3)
spm.limit.robust(t2_values, ncomp = 3, type = "t2", method = "empirical")
#> $ucl
#> [1] 5.587954
#> 
#> $alpha
#> [1] 0.05
#> 
#> $description
#> [1] "T2 empirical quantile, alpha=0.05"
#> 
#> $method
#> [1] "empirical"
#> 
spm.limit.robust(t2_values, ncomp = 3, type = "t2", method = "bootstrap")
#> $ucl
#> [1] 5.587954
#> 
#> $alpha
#> [1] 0.05
#> 
#> $description
#> [1] "T2 bootstrap (1000 resamples), alpha=0.05"
#> 
#> $method
#> [1] "bootstrap"
#> 
# }