摘要: Abstract A formalized method of constructing a best-fitting smooth curve, with confidence limits, to sequential data on sphere is proposed and least-squares alignment two time series discussed. The procedure, developed from Gould's regression technique for angular variates (corresponding multiple analysis normally distributed variates), involves fitting cubic splines using weighted (robust) approximation. Paths apparent polar wander (APW) are defined continuously, have continuous first second derivatives always lie the sphere. cross-validation mean square deviation used determine degree smoothing. North American European APW paths, their 95% bands accelerations last 500 Ma been calculated.