Sign tests for weak principal directions (Q2203629)

In the context of principal component analysis (PCA), the authors consider inference on the first principal direction of a \(p\)-variate elliptical distribution focussing on sign-based statistical procedures, that is, on procedures that involve the observations only through their direction from the center of the distribution. The null hypothesis is that the first principal direction coincides with a given direction of \(\mathbb{R}^p\). The authors propose a sign test and demonstrate its asymptotic properties. They also compare the test to some competitors. The practical relevance of the proposed sign test is illustrated on the Swiss banknote dataset, that is available in the \texttt{R} package \texttt{uskewfactors.}