Page 389 - Special Topic Session (STS) - Volume 3
P. 389
STS551 Zamira Hasanah Zamzuri et al.
3. Results
We will present the results on fitting the proposed (APL) and zero inflated
Poisson (ZIP) models to simulated data sets based on both distributions.
Hence, the results will be arranged in this order:
1) Comparison of APL and ZIP models fitted to simulated data based on
APL
2) Comparison of APL and ZIP models fitted to simulated data based on
ZIP
A detailed result will be included later.
4. Discussion and Conclusion
In this paper, we have introduced a new model for traffic accident count
data based on the adjustment to the MPL model in which a proportion
parameter is incorporated into the model. The model fitting process consists
of two phases: first estimating the proportion parameter and then estimating
other parameters in the new model using an MCMC procedure. We expect
that the simulation results reveal that this model performs better and able to
explain the presence of extra zeros due to the underreporting scenario.
It is important to note that the aim of this paper is to offer an alternative on
fitting count models to the data with extra zeros, specifically in traffic accident
environment. Furthermore, more information is obtained through the
estimation of the proportion of the reported accidents. When the proportion
of the reported accidents and reported accidents rate are estimated, this
allows us to estimate the true accident rate in which help in terms of estimation
accuracy on the actual accident count that happen.
The work here illustrates the need on reporting traffic accidents and a proper
data documentation practice. Such issue may need to be handled in the long
run, but with the current situation, especially in Malaysia; in which the source
378 | I S I W S C 2 0 1 9