A resistant measure of heteroskedasticity in explorative time series analysis

The stationarity of time series is often reached through the transformation of the observed data. When the analysis of the series is carried out automatically using implemented softwares, it is needed to define some indicators which alerts the system about the non stationarity of the data and leads to right transformations. In this context, the present paper proposes an indicator which detects the heteroskedasticity of the data and its empirical distribution has been investigated through Monte Carlo simulations. The performance of the indicator has been compared to well know homoskedasticity test usually implemented in statistical softwares.