Files in this item
|(no description provided)|
|Title:||A Time Series Approach to the Prediction of Oil Discoveries|
|Author(s):||Cartwright, Phillip August|
|Department / Program:||Economics|
|Degree Granting Institution:||University of Illinois at Urbana-Champaign|
|Abstract:||This research develops a time series methodology for the prediction of oil discoveries at the level of the petroleum play. The study develops a Bayesian, multi-state (outlier/no-outlier), time varying coefficients, nonlinear time series forecasting model and applies the model to the problem of predicting oil discoveries in the North Sea and in the region of the Niger Delta.
The results of the application of the Bayesian model to the chronologically ordered sequence of North Sea discoveries indicate that the forecasting performance of the Bayesian model dominates that of a non-Bayesian specification with respect to one-step forecast error measured over the last 20 observations. The performance of the model over the last 20 discoveries is considered to be a reasonable basis for judging model performance over the recent past. Long-range forecasts are generated for the North Sea and the results are compared with alternative assessments.
The model is applied to a chronologically ordered sequence of oil discoveries in the region of the Niger Delta. There are missing observations in the data sequence, however, and a procedure for forecasting missing observations is indicated. The model is fit and applied to generate one and ten-step forecasts. While the forecasts appear reasonable based upon available information, reliability of the forecasts is uncertain due to the presence of missing values and to the fact that the producing region did not enter maturity over the sequence of data.
This research provides support for the argument that there are likely to be substantial gains to be obtained as the result of consideration of broader classes of time series models beyond those belonging to the conventional ARMA class. More complex models are likely to provide a better fit to the data than that given by standard linear models, and the gains in modelling flexibility are likely to be substantial.
Thesis (Ph.D.)--University of Illinois at Urbana-Champaign, 1982.
|Date Available in IDEALS:||2014-12-16|