Monte Carlo Algorithm And Data Mining In Linear Modeling

The theory of linear model is an ancient branch of statistics, but for quite a long time, linear model is always a focus for statistics , even for economist. A lot of new results can be found in this field. These results can be defined as such two sides: one is the improvement of linear modeling, the other is that in some concerned research field, some complex problems were turned into linear model, because the theory of linear model is so perfect, those complex problems were simplified.This paper focus on two field: one is the non-parametric monte carlo testing in multi-linear model; the other is the testing in financial time series. Owing to the internal relations, the data in financial market usually manifest as the interrelated time series. This paper mainly discusses how to simplify time series models in financial market into relevant linear models and how to examine the existence of outliers and differentiate innovation outliers from additive outliers with traditional linear models. The mining of innovation outliers has not only the theoretical significance but also a great practical significance in the research on financial risk. Besides, the two algorithms proposed in this paper are analyzed with authentic proofs; in this way, the two methods in the study of financial market are proved feasible and effective.