Advances in Signal Extraction and Forecasting

It's been a while since I posted on the topic... This blog is the new site -- the new home for MDFA and related topics -- where I will regularly publish/share  documents, notes, publications, insights, experience, R-code on 'Signal Extraction and Forecasting' (SEF in short).

Application Fields

I'll emphasize noisy environments typically found in economic applications:

• macro series, financial series;  
• dynamics close to random-walks (levels/prices) or noise (differences/log-returns); 

and (most of) the proposed methods will be shaped - customized - in view of tackling these challenging application fileds.

 

Linear, Non-Linear and Hybrid

Not so long ago, I was inclined to make a clear-cut distinction between linear (for example classic time series analysis) and non-linear approaches (for example neural networks); but since I intend to report on hybrid designs (combining classic time series analysis with neural networks, for example), I suggest to think in terms of a 'solution of continuity' linking smoothly both paradigms.

Starting Point

I thought that the following presentation Advances in Signal Extraction and Forecasting (left-click on the link) would be a good starting point.

• This collection of slides addresses SEF-topics potentially relevant to quant/research-teams in the finance industry. 
• It provides an overview of the modular MDFA-body with a short motivation for each module (the practical problem, the proposed solution and links to examples and results addressed by each MDFA 'plug-in') 

I'm currently in the process of collecting feedback/experience with these slides (touring in the industry); as an outcome of that on-going feedback process I decided to start this blog.

What's next?

I intend to go through the above slides in a series of dedicated posts that should help in highlighting (some of) the key methodological and practical SEF-issues. A first tutorial emphasizing multivariate signal extraction in a mean-square error framework is provided in 1.

• It is designed as an entry-point to the methodology discussed in the above slides as well as to the  MDFA-package deposited on Github, see 2.  
• It replicates one of the slides in the above presentation...

What Happened with (the now Old/Obsolete) SEFBlog?

In a nutshell: I lost confidence in the support/stability provided by the previous environment (I endured loss of most code/figures/material after a chaotic imposed migration process). The current private framework is more flexible and 'fully' under (my) control. Is it more stable/safe/reliable? I don't know. But at least one exogeneous shock-supplier has been eliminated in moving the house.