Print a Nice "summary" of Robustness Weights

Print a nice “summary” about a numeric vector of robustness weights. Observations with weights around zero are marked as outliers.

summarizeRobWeights(w, digits = getOption("digits"),
                    header = "Robustness weights:",
               eps = 0.1 / length(w), eps1 = 1e-3, ...)

Arguments

w

numeric vector of robustness weigths.

digits

digits to be used for printing.

header

string to be printed as header line.

eps

numeric tolerance \(\epsilon\): values of w with \(\left|w_i\right| < \epsilon/n\) are said to be outliers.

eps1

numeric tolerance: values of w with \(\left|1 - w_i\right| < eps1\) are said to have weight ‘~= 1’.

...

potential further arguments, passed to print().

See also

The summary methods for lmrob and glmrob make use of summarizeRobWeights().

Our methods for weights(), weights.lmrob(*, type="robustness") and weights.glmrob(*, type="robustness").

Value

none; the function is used for its side effect of printing.

Author

Martin Maechler

Examples

w <- c(1,1,1,1,0,1,1,1,1,0,1,1,.9999,.99999, .5,.6,1e-12)
summarizeRobWeights(w) # two outside ~= {0,1}
#> Robustness weights: 
#>  3 observations c(5,10,17) are outliers with |weight| <= 1e-12 ( < 0.005882); 
#>  12 weights are ~= 1. The remaining 2 ones are
#>  15  16 
#> 0.5 0.6 
summarizeRobWeights(w, eps1 = 5e-5)# now three outside {0,1}
#> Robustness weights: 
#>  3 observations c(5,10,17) are outliers with |weight| <= 1e-12 ( < 0.005882); 
#>  11 weights are ~= 1. The remaining 3 ones are
#>     13     15     16 
#> 0.9999 0.5000 0.6000 

## See the summary(<lmrob>) outputs