This final degree was an opportunity for me to investigate in one of the most /ninteresting fields from economics: the analysis of real financial data with statistics for /nthe understanding of the stock market. The methods of data analysis, specifically/neconometrics, statistics or data analysis have evolved exponentially during these last /ndecades because of the recent availability of huge amounts of information. Thanks to /ncomputer science and the huge amounts of available data (which is also ...
This final degree was an opportunity for me to investigate in one of the most /ninteresting fields from economics: the analysis of real financial data with statistics for /nthe understanding of the stock market. The methods of data analysis, specifically/neconometrics, statistics or data analysis have evolved exponentially during these last /ndecades because of the recent availability of huge amounts of information. Thanks to /ncomputer science and the huge amounts of available data (which is also called Big /nData), the evaluation of these large quantity of information and the extraction of /nmeaningful results can be accomplished cost effectively. This assessment given by data analysts provides valuable information to public and private companies to better /nmanage their resources. The scope of this final degree project is to learn how to use /none of the tools used in Big Data analysis as well as applying it to the project: Partial /nCorrelation Networks with a multivariate time series analyzed with the LASSO /nestimation technique./nFirstly, the features of the instruments used in the project are explained. Specifically, /nthe statistical models and tools used to produce illustrated network graph and tables /nthat displays the correlations among different elements of the data set. This has also /nbeen performed with Garch (1,1), which is a model that fits quite well financial data. /nFinally, empirical data analysis has been made in order to conclude the project with /nsome findings that can give interesting results for the understanding of the data set/nused./nThe project results indicate the Partial Correlation networks by Joint Sparse /nRegressions Models approach presented by Peng et al. (2009) is an effective analysis /ntool. This method does assume data sparsity.. Additionally, conditional temporary /ndependence across different variables in time has proved to work well under different /ndata sets, and also in the data from this project. /nOut of the many tools and techniques available, this project implemented only the /nLASSO estimation technique in partial correlation networks using returns on financial /ndata and Garch residuals. For this reason, the scope of this project is limited. There /ndifferent areas and analysis that could be done to the same data, so that it has just /nbeen taken some of them that are considered relevant. Also, the choice of giving /nvalues to the parameter lambda is also limited since it has been analyzed only eight of /nthem. For these reasons, the project is limited and deeper analysis on the data could /nbe done in order to be more complete./nThe purpose of this project is to report an analysis about a multivariate time series /ndata set and understanding how the tools described above are used and implemented.
+