Degree Type

Creative Component

Semester of Graduation

Fall 2018

Department

Statistics

First Major Professor

Farzad Sabzikar

Degree(s)

Master of Science (MS)

Major(s)

Statistics

Abstract

The objective of this creative component is to learn the ARTFIMA time series model and use it to fit real world data and compare the model with ARFIMA model in terms of goodness of fit and predictions. The ARTFIMA model is a short range dependent time series that exhibits semi-long range dependency. For small values of tempering parameter λ > 0, the spectral density behaves like a power law at low frequencies, and it remains bounded as frequency reaches to zero. ARTFIMA can be extended to Autoregressive Fractionally Integrated Moving Average (ARFIMA) when the tempered paramater λ is zero, and to Autoregressive Moving Average (ARMA, which is most commonly used time series model) when both tempered parameter λ and difference parameter d are zero. We can consider ARTFIMA model as a generalized time series model, where as both ARFIMA and ARMA are the extensions of ARTFIMA.

Copyright Owner

Dinesh Reddy Poddaturi

File Format

application/pdf

Share

COinS