Unification of Statistical Methods for Continuous and Discrete Data

Based on BP Neural Network Discrete Data Forecast

Applied Mechanics and Materials ◽

10.4028/www.scientific.net/amm.50-51.977 ◽

2011 ◽

Vol 50-51 ◽

pp. 977-981 ◽

Cited By ~ 2

Author(s):

Jing Wang ◽

Guo Li Wang ◽

Jian Hui Wu ◽

Yu Su

Keyword(s):

Neural Network ◽

Artificial Neural Network ◽

Statistical Methods ◽

Bp Neural Network ◽

Model Fitting ◽

Predictive Ability ◽

Discrete Data ◽

Engineering System ◽

Forecast Performance ◽

Artificial Neural

Artificial neural network is based on human brain structure and operational mechanism based on knowledge and understanding of its structure and behavior of simulated an engineering system. BP artificial neural network is an important component of neural networks, as it can on the linear or nonlinear multivariable without preconditions in the case of statistical analysis, with the traditional statistical methods, analysis of the variables need to be consistent with certain conditions compared to its own advantage. The BP neural network does not need the precise mathematical model, does not have any supposition request to the material itself. Its processing non-linear problem's ability is stronger than traditional statistical methods. This article uses two groups of data to establish the BP neural network model separately, and carries on the comparison to the model fitting ability and the forecast performance, discovered BP neural network when data distribution relative centralism fits ability, forecasts the stable property. But the predictive ability is unable in the discrete data application to achieve anticipated ideally.

Download Full-text

Unification of Statistical Methods for Continuous and Discrete Data

Computing Science and Statistics ◽

10.1007/978-1-4612-2856-1_30 ◽

1992 ◽

pp. 235-242

Author(s):

Emanuel Parzen

Keyword(s):

Statistical Methods ◽

Discrete Data

Download Full-text

Integrated Fractional white Noise as an Alternative to Multifractional Brownian Motion

Journal of Applied Probability ◽

10.1017/s0021900200003041 ◽

2007 ◽

Vol 44 (02) ◽

pp. 393-408 ◽

Cited By ~ 1

Author(s):

Allan Sly

Keyword(s):

Stochastic Process ◽

Brownian Motion ◽

White Noise ◽

Gaussian Process ◽

Discrete Data ◽

Hölder Exponent ◽

Scaling Properties ◽

Multifractional Brownian Motion ◽

Local Properties

Multifractional Brownian motion is a Gaussian process which has changing scaling properties generated by varying the local Hölder exponent. We show that multifractional Brownian motion is very sensitive to changes in the selected Hölder exponent and has extreme changes in magnitude. We suggest an alternative stochastic process, called integrated fractional white noise, which retains the important local properties but avoids the undesirable oscillations in magnitude. We also show how the Hölder exponent can be estimated locally from discrete data in this model.

Download Full-text

Statistics and parallaxes

International Astronomical Union Colloquium ◽

10.1017/s0252921100073875 ◽

1978 ◽

Vol 48 ◽

pp. 7-29

Author(s):

T. E. Lutz

Keyword(s):

Experimental Design ◽

Statistical Methods ◽

Review Paper ◽

Systematic Errors ◽

Past Experience ◽

Random Errors ◽

Trigonometric Parallaxes ◽

Systematic And Random Errors

This review paper deals with the use of statistical methods to evaluate systematic and random errors associated with trigonometric parallaxes. First, systematic errors which arise when using trigonometric parallaxes to calibrate luminosity systems are discussed. Next, determination of the external errors of parallax measurement are reviewed. Observatory corrections are discussed. Schilt’s point, that as the causes of these systematic differences between observatories are not known the computed corrections can not be applied appropriately, is emphasized. However, modern parallax work is sufficiently accurate that it is necessary to determine observatory corrections if full use is to be made of the potential precision of the data. To this end, it is suggested that a prior experimental design is required. Past experience has shown that accidental overlap of observing programs will not suffice to determine observatory corrections which are meaningful.

Download Full-text