27.08.2019 Views

book

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Poster Communications<br />

SP1<br />

Wednesday, September 4th<br />

19:45<br />

A new autorregressive moving average model<br />

based on the unit Rayleigh distribution for rates<br />

and proportions: a simulation study<br />

Renata Rojas Guerra<br />

UFSM<br />

Fernando A. Peña-Ramírez<br />

The unit Rayleigh (UR) is a one-parameter model that is suitable to accommodate asymmetric<br />

unimodal data. It was pioneered Mazucheli et. al. (2018) as a special case of the unit Weibull<br />

distribution. From the results of Guerra (2019), we can cite two main advantages of this model.<br />

The first one is that it presents the maximum likelihood estimator in closed-form. Secondly, it<br />

has a simple expression for the median. In this work, we consider a median re-parametrization<br />

for the UR distribution and propose an autoregressive moving average structure based on this.<br />

We aim to provide a simple alternative for modeling double bounded variables under the presence<br />

of serial correlation in the conditional median of the UR distribution. Once the parent model is<br />

a one-parameter distribution, the introduced model is more parsimonious than the most popular<br />

autoregressive moving average time series. We discuss a conditional maximum likelihood approach<br />

to estimate the model parameters and present a Monte Carlo simulation study to assess its finite<br />

sample performance.<br />

Keywords: ARMA models; Bounded data; Unit Rayleigh distribution<br />

56

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!