STATISTICAL TESTS OF STOCHASTIC-PROCESS MODELS USED IN THE FINANCIAL THEORY OF INSURANCE COMPANIES

This paper presents a statistical test allowing the analyst to determi ne if a given time series is statistically incompatible with being mod eled as a linear or log-linear process. Since the commonly used models for financial time series of interest to insurance professionals are linear or log-linear, this paper allows the analyst to verify the line arity of the model under investigation, or else points to the necessit y of non-linear modeling. We also show how to test for time series Gau ssianity using the same type of statistical test statistic. These resu lts are applied to several financial data sets relevant to the financi al operations of insurance companies.