qn: Robust scale estimate Qn
In floatofmath/adaperm: Permutation tests for adaptive designs

Description Usage Arguments Details Value Author(s)

Compute the robust scale measure Qn given in "Rousseeuw and Croux (1993)" which is about 86

1	qn(x, factor = 2.2219)

`x`	vector of observations
`factor`	Scaling factor

The default scaling factor is chosen such that the estimate is approximately consistent for normal observations. For different reference distributions other factors may be more apropriate. Note however, that we have hardcoded the bias correction constants applied to small samples - these may differ as well for other reference distributions. Changing the factor may be futile.

Note that we use a computationally suboptimal implementation that uses O(n^2) operations. If time allows we may implement the more efficient method given in "Croux and Rousseeuw (1992)" which requires only O(n*log(n)) operations.

numeric value

Florian Klinglmueller

floatofmath/adaperm documentation built on May 16, 2019, 1:18 p.m.