We show that Kolmogorov complexity and such its estimators as

universal codes (or data compression methods) can be applied for

hypothesis testing in a framework of classical mathematical

statistics. The methods for identity testing and nonparametric

testing of serial independence for time series are described.