Select Data Type:

Equal Weights (Default) Unequal Weights Error Bars

Format: One value per line. Weights, if unequal, should be included on the same line, separated by white space.

Alternatively, (symmetric) error bars may be entered instead of weights. These will be converted to weights using weight = 1/error². However, if the scatter of the data is too large to be accounted for by the error bars alone, extrinsic scatter dominates and "Equal Weights" should be selected instead (§8.2.1, Maples et al. 2018).

Select Uncontaminated Distribution:

Gaussian, Mildly Peaked, or Mildly Flat-Topped (Default) Mildly Asymmetric (Less Powerful) Other

Select Contaminant Type:

Selected Uncontaminated Distribution:

Selected Rejection Method:

μ: Median (§2.1; Maples et al. 2018)
σ : Broken-line technique (§2.2, §2.3)
Rejection proportional to σ both below and above μ (§3.1, §3.3.2)

μ : Half-sample mode (§2.1; Maples et al. 2018)
σ_- , σ₊ : 68.3%-Value technique (§2.2, §2.3)
Rejection proportional to min{σ_- , σ₊} both below and above μ (§3.2, §3.3.2)

μ: Half-sample mode (§2.1; Maples et al. 2018)
σ_- , σ₊ : Broken-line technique (§2.2, §2.3)
Rejection proportional to min{σ_- , σ₊} both below and above μ (§3.2, §3.3.2)

μ: Half-sample mode (§2.1; Maples et al. 2018)
σ_- , σ₊ : Broken-line technique (§2.2, §2.3)
Rejection proportional to σ_- below μ and σ₊ above μ (§3.3.1)

Fundamentally, RCR assumes that the uncontaminated distribution is Gaussian, or near Gaussian. In particular, our All High or All Low and In Between contaminant type options are especially sensitive to this assumption, and should not be used if this assumption is not met.

Note: Rejection method is iterative, preceded by iterative bulk rejection (§5), and followed by both (1) iterative "median + 68.3%-value technique" rejection, and (2) iterative "mean + standard deviation " rejection (§4).