B0334
Title: Statistical inference in structural equation modeling with latent variables for diffusion processes
Authors: Shogo Kusano - Osaka University (Japan) [presenting]
Masayuki Uchida - Osaka University (Japan)
Abstract: Structural equation modelling (SEM) is considered with latent variables for diffusion processes. SEM is a statistical method that describes the relationships between latent and observable variables. While many researchers have studied SEM for IID models, there are few studies on SEM for time series models. Recently, since high-frequency data can be easily obtained, such as stock price data, statistical inference for diffusion processes based on high-frequency data has been extensively researched. Therefore, SEM is proposed for diffusion processes based on high-frequency data. The quasi-likelihood estimators for parameters in the SEM are obtained. The goodness-of-fit test is also introduced using the quasi-likelihood ratio. It is shown that proposed statistics have good asymptotic properties. Furthermore, some examples and simulation studies of the proposed statistics are given.
