Statistical assessment of the potential factors affecting delayed incident reporting in the oil and gas industry

Open Access
Abbe, Osahon Ighodaro
Graduate Program:
Energy and Mineral Engineering
Master of Science
Document Type:
Master Thesis
Date of Defense:
April 02, 2012
Committee Members:
  • Samuel Oyewole, Thesis Advisor
  • Safety management
  • decision-making
  • analytical hierarchy process
  • oil and gas
  • delayed incidents
  • regression analysis
Safety in the workplace has an ever-growing audience; even so safety in the workplace in the oil and gas industry. The Occupational Safety and Health Administration (OSHA) enforces and regulates safety rules through federal-approved and state-approved programs that must be met by businesses. In this thesis, safety records from an undisclosed company in the oil and gas industry were used to study factors that potentially contribute to having an effect on the health of humans in areas where there have been oil spills. The safety records from the study included chemical spills, explosions and safety incidents over a 21-year period. The objective of this research was to develop the most feasible model to help in identifying the significant independent factors such as nature of spill, delay time, incident type, population, yearly effect and seasonal effect in delayed incident reporting. In this study, the various models were developed based on the previous spills which can be used to predict potential health effects of future spills and effectively manage the effects or eliminate them. A dependent variable (deaths per state) was regressed against the significant independent factors which were converted from qualitative to quantitative data points using analytical hierarchy process (AHP) technique. Additional transformational analyses were performed using the square root, cube, square, inverse and natural logarithm of the dependent variable. From the transformational analyses, the most feasible model was the cube transformation model of the dependent variable which had the highest R-squared value while still maintaining its normality. The cube transformation of the dependent variable had an F-value of 985.13 and R-square value of 79.6% compared to the base model which had an F-value of 375.14 and R-square value of 51.9%. The cube model used the independent factors to predict the potential health effects from the dependent variables and the significant factors that proved to be helpful in reducing the negative effects of the incidents were the delayed reporting and the nature of the chemical spilled. However, with the data set lacking some critical information pertaining to the corresponding injuries and economical cost per incident, adequate analyses on the direct impact of each incident was limited.