<div dir="ltr"><br><div class="gmail_quote"><br><div style="font-size:10pt;font-family:Verdana,Geneva,sans-serif">
<p class="m_1264526126443114936western" style="text-align:center"><font size="6"><b><span>Universidade de Brasília</span></b></font></p>
<p class="m_1264526126443114936western"> </p>
<p class="m_1264526126443114936western" style="text-align:left" align="center"><span style="font-size:x-large"><b><span style="font-size:14pt">O Programa de Pós-Graduação em Estatística convida para</span>:</b></span></p>
<p class="m_1264526126443114936western" align="center"> </p><span style="color:#ffffff"><span style="font-size:300%"><b><span>ALESTRA</span></b></span></span>
<p class="m_1264526126443114936western" align="center"><b><span><span style="color:#000000"><span style="font-family:'Times New Roman',serif"><span style="font-size:xx-large"><span><span><span><u>Time series forecasting, model competitions and the theta method</u></span></span></span></span></span></span></span></b></p>
<p class="m_1264526126443114936western" align="center"> </p>
<p class="m_1264526126443114936western" style="text-align:left" align="center"><span style="font-size:12pt"><b>Palestrante:  </b></span><span style="font-size:12pt;font-family:arial,helvetica,sans-serif"><span><span><span>Prof. </span></span></span><span><span><span>José</span></span></span><span><span><span> Augusto Fiorucci</span></span></span> <span><span> (UnB </span></span><span><span><span>EST</span></span></span><span><span>)</span></span></span></p>
<p class="m_1264526126443114936western"><span style="font-size:12pt"><span><b>DATA: </b></span><span style="font-family:arial,helvetica,sans-serif"><span style="color:#000000"><span><span>1</span></span></span><span style="color:#000000"><span><span>0</span></span></span><span style="color:#000000"><span>/0</span></span><span style="color:#000000"><span>5</span></span><span style="color:#000000"><span>/2018</span></span><span style="color:#000000"><span> </span></span><span>(</span><span style="color:#000000"><span><span>quinta</span></span></span><span>-feira)</span></span></span></p>
<p class="m_1264526126443114936western"><span style="font-size:12pt"><span><b>HORÁRIO: </b></span><span>14:30h</span></span></p>
<p class="m_1264526126443114936western"><span style="font-size:12pt"><span><b>LOCAL: </b></span><span style="color:#000000;font-family:arial,helvetica,sans-serif"><span>Sala Multiuso (A1-7/76)</span></span></span></p>
<p class="m_1264526126443114936western" align="center"><span style="font-family:'Times New Roman',serif"><span style="font-size:xx-large"><b>Resumo</b></span></span></p>
<p class="m_1264526126443114936western" align="center"> </p>
<p class="m_1264526126443114936western" align="justify"><span style="font-family:'Times New Roman',serif"><span style="font-size:medium">I<span style="font-size:medium">n this talk we will present recent stochastic expansions of the theta method, the winner of the largest time series competitions (M3-Competion), which caught researchers' attention due to its performance and simplicity. The Theta method, as implemented in the monthly subset of the M3-Competiton, decomposes the seasonally adjusted data into two "theta lines''. The first theta line removes the curvature of the data to estimate the long-term trend component. The second theta line doubles the local curvatures of the series to approximate the short-term behavior. The proposed Dynamic Optimised Theta Model is a state space model that optimally selects the best short-term theta line and dynamically revises the long-term theta line. The superior performance of this model is demonstrated through an empirical application in the M3-Competition database.<br> <br> Referências:<br> Makridakis, S. & Hibon, M. (2000). The M3-competition: results, conclusions and implications. International Journal of Forecasting , 16, 451{476.<br> Fiorucci, J. A., Pellegrini, T. R., Louzada, F., Petropoulos, F. & Koehler, A. B. (2016). Modelfor optimising the theta method and their relationship to state space models. Internationa Journal of Forecasting, 32(4), 1151-1161.</span></span></span></p>
<p class="m_1264526126443114936western" align="justify"><span style="font-family:'Times New Roman',serif"><span style="font-size:medium">Fiorucci, J. & Louzada, F. (2016). forecTheta: Forecasting Time Series by Theta Models, R package version 2.2, CRAN.</span></span></p><span class="HOEnZb"><font color="#888888">
<div>-- <br>
<div class="m_1264526126443114936pre" style="margin:0;padding:0;font-family:monospace">Programa de Pós-Graduação em Estatística<br> Universidade de Brasília<br> <a href="http://www.pgest.unb.br" rel="noreferrer" target="_blank">www.pgest.unb.br</a><br> Telefone: (61) 3107-3697</div>
</div>
</font></span></div>
</div><br></div>