In this study, a count-data regression is presented
to estimate and analyze the effects of determinant factors affecting the
accidents leading to death, through negative binomial regression. For
this purpose the structure of 50 accidents that led to death and another
2700 accidents in the Construction Phase in Oil, Gas and Petrochemical
Projects (a case study of Assaloyeh) during 2003-2005 has been studied.
Along with total accidents, unsafe conditions, human errors, management
faults and using nonstandard equipments, were considered as the main independent
variables affecting the job accidents leading to death, as the dependent
variable. By employing the method of developing abstract variables and
taking values (codes) one and zero (zero for lack of quality and one for
its existence), the variables were quantified. EViews software has been
employed, because it provides support for the estimation of several models
of count data. The findings of the study show that for each number increase
in the unsafe conditions, human errors and either nonstandard equipments
or management faults, the expected number of deadly accident increases
by a factor of 0.2982 and 0.1137 as well as 0.0259, respectively. If the
number of total accidents increases by one unit, the difference in the
logs of expected counts would be expected to increase by 0.0025 unit,
while holding the other variables in the model constant. Apart from such
predictors, the log of the expected count for deadly accidents is 0.0023.
Seyed Bagher Mortazavi, Abbas Zarae Nezhad, Mansour Zarra Nezhad and Hasan Asilian Mahabadi, 2007. Examining the Determinants of Occurrence of Accidents at the Construction
Phase in Oil, Gas and Petrochemical Projects: (A Case Study of Assaloyeh). Journal of Applied Sciences, 7: 1088-1092.