Chain Error as a function of Seasonal Variation

Abstract

In this study, we examine statistically the dependence between Seasonal Variation of consumed values and the ChainErrors of corresponding excellent indices in different subgroups Ak.

First, cyclic seasonal variation of values is calculated by simple regression analysis and the ChainError is calculated by the Multi Period Identity Test. Secondly, Quadratic Means QM of these two variables (or dimensions) are used in our analysis. Question is: Does the largeness of the seasonal components in the value series, as measured by its Quadratic Mean (QM) per month during the observation period, reflect itself in the largeness of ChainErrors (CE) derived by Multi Period Identity Test?

The Quadratic Means of cyclic seasonal variation of values and ChainError (difference between base and chain strategies) both show variation found in typical average months. The dependence between these two quadratic means is shown in the paper by simple regression analysis. We show that there is a very strong statistically significant dependency between Quadratic Means of Chain Errors and Quadratic Means of values in the seasonal index. Our main empirical findings are following: Do not use any construction strategy that is somehow connected with the chain strategy.

Our test data is a scanner data from one of Finnish retail trade chains including monthly information on unit prices, quantities and values from January 2014 to December 2018, and has more than 20 000 homogeneous commodities that are comparable in quality.