Naoko Sakurai

Statistical Regularities of Seismic Noise

Seismic noise includes the information that comes from crust of earth and hypocenter. The study of seismic noise is an essential topic for earthquake predictability and investigation of crustal structure. In this paper we analyze the seismic noise from 46 stations in all around Japan of Full Range Seismograph Network of Japan (F-net). To investigate the influence from earthquake we analyze two types of seismic noise data, one is the data after earthquake shake, which from 2,000 seconds after the max shake of an earthquake, and others are without any earthquake effects. We find that (i) the magnitude of seismic noise values follows a log-normal distribution and (ii) the conditional mean growth rate and the standard deviation of the growth rate of the magnitude value both obey power-law with the magnitude value. Furthermore we also find that the Hurst exponent depends on the max amplitude of earthquake, and shows a linear correlation with logarithmic max amplitude.

Correlation and hierarchies in financial markets

As an emerging financial market, the trading value of carbon emission trading market has definitely increased in recent years. The carbon emission is not only trading in carbon emitters but also has become an important investment target. To determine the mechanism of this growing market, we analyzed the EU allowances (EUA) price series in European Climate Exchange (ECX) that is the leading European emissions futures market. As other financial market, the absolute value of price change (volatility) in carbon emission trading market also shows long-term power-law correlations. Our analysis shows that definite cross correlations exist between EUA and many other markets. These cross correlations exist in wild-range fields, stock market index, futures of crude, sugar, cocoa, etc., suggesting that in this new carbon emission trading market the speculation behavior had already become a main factor that can affect the price change.

Latent class model with two latent variables for analysis of count data

Many latent class models have been proposed for analysis of recurrent count and choice data. In this paper, a latent class model with with two latent variables for analysis of count data. Examples of analysis of real data are also given.