<div dir="ltr"><p class="MsoNormal" align="center" style="margin-bottom:0.0001pt;text-align:center;line-height:normal"><b><span style="font-size:14pt;font-family:"Times New Roman",serif">Ciclo de Palestras online do Departamento de Estatística da UFBA</span></b><span style="font-size:14pt;font-family:"Times New Roman",serif"></span></p><p class="MsoNormal" style="margin-bottom:0.0001pt;line-height:normal"><span style="font-size:14pt;font-family:"Times New Roman",serif"> </span></p><p class="MsoNormal" style="margin-bottom:0.0001pt;line-height:normal"><b><span style="font-size:14pt;font-family:"Times New Roman",serif">Palestrante</span></b><span style="font-size:14pt;font-family:"Times New Roman",serif">: Hedibert Freitas Lopes (INSPER)</span></p><p class="MsoNormal" style="margin-bottom:0.0001pt;line-height:normal"><span style="font-size:14pt;font-family:"Times New Roman",serif"> </span></p><p class="MsoNormal" style="margin-bottom:0.0001pt;line-height:normal"><b><span lang="EN-US" style="font-size:14pt;font-family:"Times New Roman",serif">Title</span></b><span lang="EN-US" style="font-size:14pt;font-family:"Times New Roman",serif">: Sparse and Time-varying Covariance Modeling</span></p><p class="MsoNormal" style="margin-bottom:0.0001pt;line-height:normal"><span lang="EN-US" style="font-size:14pt;font-family:"Times New Roman",serif">CV: </span><span style="font-size:14pt;font-family:"Times New Roman",serif"><a href="http://lattes.cnpq.br/7679840055825288" target="_blank"><span lang="EN-US">http://lattes.cnpq.br/7679840055825288</span></a></span><span lang="EN-US" style="font-size:14pt;font-family:"Times New Roman",serif"></span></p><p class="MsoNormal" style="margin-bottom:0.0001pt;line-height:normal"><span lang="EN-US" style="font-size:14pt;font-family:"Times New Roman",serif"> </span></p><p class="MsoNormal" style="margin-bottom:0.0001pt;line-height:normal"><span lang="EN-US" style="font-size:14pt;font-family:"Times New Roman",serif">Abstract: In many areas there has been a growing interest in developing methods that can estimate dependency of multivariate time series data. Nevertheless, the estimation of a high-dimensional covariance matrix which potentially varies over time is still an open problem. In this work, we discuss several bayesian regularization methods based on shrinkage and variable selection priors and estimate sparse covariance matrices using the modified Cholesky decomposition. Our first application considers time-varying observational variances (stochastic volatility) and uses the Normal-Gamma prior for shrinking the regression coefficients that compose the Cholesky factor. The second application considers homoscedastic errors and time varying regression coefficients which are shrunken by applying an extension of the Normal Mixture of Inverse Gammas (NMIG) hierarchical prior for dynamic models, which accommodates time varying sparsity.</span></p><p class="MsoNormal" style="margin-bottom:0.0001pt;line-height:normal"><span lang="EN-US" style="font-size:14pt;font-family:"Times New Roman",serif"> </span></p><p class="MsoNormal" style="margin-bottom:0.0001pt;line-height:normal"><b><span style="font-size:14pt;font-family:"Times New Roman",serif">Data</span></b><span style="font-size:14pt;font-family:"Times New Roman",serif">: 25/08/2017                                                        </span></p><p class="MsoNormal" style="margin-bottom:0.0001pt;line-height:normal"><b><span style="font-size:14pt;font-family:"Times New Roman",serif"> Hora</span></b><span style="font-size:14pt;font-family:"Times New Roman",serif">: 11hs</span></p><p class="MsoNormal" style="margin-bottom:0.0001pt;line-height:normal"><b><span style="font-size:14pt;font-family:"Times New Roman",serif">Local</span></b><span style="font-size:14pt;font-family:"Times New Roman",serif">: Auditório do IME para aqueles que desejam acompanhar de forma presencial.</span></p><p class="MsoNormal" style="margin-bottom:0.0001pt;line-height:normal"><span style="font-size:14pt;font-family:"Times New Roman",serif"> </span></p><p class="MsoNormal" style="margin-bottom:0.0001pt;line-height:normal"><b><span style="font-size:14pt;font-family:"Times New Roman",serif">O link para acompanhar a palestra é</span></b><span style="font-size:14pt;font-family:"Times New Roman",serif"></span></p><p class="MsoNormal" style="margin-bottom:0.0001pt;line-height:normal"><span style="font-size:14pt;font-family:"Times New Roman",serif"><a href="https://www.youtube.com/channel/UCC96Rmc3qKEYkKk187IcLdA/live" target="_blank">https://www.youtube.com/channel/UCC96Rmc3qKEYkKk187IcLdA/live</a></span></p><p class="MsoNormal" style="margin-bottom:0.0001pt;line-height:normal"><span style="font-size:14pt;font-family:"Times New Roman",serif">E todos que estiverem acompanhando de forma online poderão enviar dúvidas pelo chat da transmissão, pois as perguntas serão repassadas para o Prof. Hedibert Freitas Lopes.</span></p><p class="MsoNormal" style="margin-bottom:0.0001pt;line-height:normal"><span style="font-size:14pt;font-family:"Times New Roman",serif"> </span></p><p class="MsoNormal" style="margin-bottom:0.0001pt;line-height:normal"><span style="font-size:14pt;font-family:"Times New Roman",serif"> </span></p><p class="MsoNormal" style="margin-bottom:0.0001pt;line-height:normal"><span style="font-size:14pt;font-family:"Times New Roman",serif">Essa atividade tem coordenação de Lizandra C. Fabio (<a href="mailto:lizandra.fabio@ufba.br" target="_blank">lizandra.fabio@ufba.br</a>) e Bruno Santos (<a href="mailto:brunorsantos@ufba.br" target="_blank">brunorsantos@ufba.br</a>), caso queiram enviar comentários ou sugestões sobre as palestras.</span></p><div><span style="font-size:14pt;font-family:"Times New Roman",serif"><br></span></div></div>