journal article
LitStream Collection
Some Vector and Arithmetic Operations on Two-Dimensional Orientation Variates, with Applications to Geological Data
doi: 10.1086/626391pmid: N/A
In part I we compare vector ($$\bar{a}$$) and arithmetic ($$\bar{x}$$) means of circular-type variates by showing how each changes with addition of one measurement (an) to (n - 1) measurements. Differentiating, $$\sigma x/\sigma a_n = 1/n$$ (no dispersion term appears here); $$\sigma a/\sigma a_n = (R_1 cos a_n + 1)/(R^2_1 + 2R_1 cos a_n + 1)$$, where R1 the vector resultant of the first (n - 1) measurements, is a measure of dispersion. As $$a_n \rightarrow 0$$ and as $$R_1 \rightarrow (n - 1)$$ (i.e., dispersion of the $$[n - 1] measurements \rightarrow 0$$), then $$\sigma a/\sigma a_n \rightarrow \sigma \bar{x}/\sigma a_n$$. Thus we obtain requirements for approximate equivalence. In part II, using trigonometric symmetry to simplify computations, a short method for computing vector mean and vector resultant is presented. We then examine ten sets of earth-science data; to nine of these we attempt to fit linear normal and circular-type normal density functions, testing goodness of fit with x2. The former (i.e., linear) fits satisfactorily in five out of nine comparisons. Circular-type frequencies are computed in four of nine examples; two of the four fit satisfactorily. In part III, we define the dispersion measure ß' as the angle enclosing a fictitious uniform distribution having the same vector strength ($$\bar{a} \equiv R/n$$) as the observed data. After showing a high linear correlation between β' and s (standard deviation) of our earth-science data, we derive the basis for this relationship in terms of equivalent uniform distributions. Further, the corollary relation between $$\bar{a}$$ and s is shown to be $$\bar{a} = (sin \sqrt{3} s) / \sqrt{3} s$$. Concluding remarks deal chiefly with this study's shortcomings.