The term mixed-frequency is used to describe data with (time) series which are not sampled at identical time-scales: for example a `mix’ of daily, weekly, monthly, or quarterly series. The resulting methodology also applies in the case of unequal release dates (for example a set of monthly indicators released at different time points in the month). The following ébauche of a working-paper summarizes possible approaches to mixed-frequency data in the framework of MDFA (i.e. a pure frequency-domain approach to the problem): mixed_frequency. A `hot’ topic would be to combine monthly macro-indicators with weekly or daily market data. Very hot…
My sabbatical just ended: it was a very productive time and (MDFA-) progresses were sustained and steady; in particular in the domain of financial trading (with MDFA). Unfortunately I won’t be able (neither willing) to poste most of these results here because it is shared/common work, subject to some (respectful) confidentiality rules. However, I will be able to poste particular technical MDFA-features, here and there (my next poste will be an esquisse of a working paper about `mixed-frequency data’). To summarize: expect on-going work of `general interest/utility’ (on MDFA and/or forecasting) to be published here.
I recently stumbled across a series of `Mathematical Challenges’ as published by the swissQuant group: here’s the latest/current March 2014 challenge. It’s about a well-known regularization of a classical (mean-square) regression estimate and some of the problems – numerical challenges – associated to it. In this context I thought it to be worthwile to contrast the proposed LASSO-method with the Regularization Troika introduced in my elements vade mecum.
A friend emailed me: "I recently had a question from somebody I met at
R/Finance about using DFA for something they were doing. Their problem
seems to be univariate, so I grabbed I-DFA out of The Book (see 1) and set up a
little script for them.What a sweet little solution that is. Fast as hell and really powerful."
In this context I’d like to take the opportunity to post a recent student-thesis (industry partner was UBS) entitled "10 Year German Government Bond Yields: Three Month Forecast with Exponential Smoothing and Direct Filter Approach":
You may skip the German blurb and go straight to the main text. The document is `sweaved': results (tables/plots) were obtained by the (R-) code as edited in the document. DFA-code is from 1: sweet, fast and powerful.
In our revised DFA-paper 1 Tucker and I added a new section (section 5.2) on seasonal adjustment. This new material originates in a talk held at the Census Bureau, last August:
The term `Direct Seasonal Adjustment’ (DSA) refers to an application of
the Direct Filter Approach (DFA) to real-time seasonal adjustment
problems. Corresponding R-code is available (on request).
Here’s the newest (revised) trilemma paper co-authored with Tucker McElroy (Census Bureau, Washington):
R-code (Sweave-environment) is provided below.
Here’s the newest (revised) DFA-paper co-authored with Tucker McElroy (Census Bureau, Washington):
R-code (Sweave-environment) is provided below. Continue reading
Today I was working on my presentation for tomorrow’s talk on DSA (Direct Seasonal Adjustment) at the Census Bureau. Late this morning I made an interesting experiment… Continue reading
When trying to look at myself on the web (that happens even to the mildest ego) I found another SEFBlog: Save Elephant Foundation. That gave me to think… (try shrinking an elephant)
I here provide some general background motivating some of my recent methodological efforts… Continue reading