I’m currently working on the customization chapter in the Legacy-project and here’s a short teaser:
- I show that customized designs outperform the best theoretical mean-square error (MSE-) filter, assuming knowledge of the true data-generating process (DGP), in terms of speed (smaller time-shift) AND noise-suppression (smoother output), both in-sample as well as out-of-sample. To be fair, this has been shown in McElroy and Wildi (ATS-trilemma paper), already, but the main `added-value’ in the book is that I reconciled different code sources i.e. results are safeguard.
- Going beyond, I show that a customized univariate filter also outperforms a bivariate MSE-design relying on an anticipative leading-indicator (lead by one time unit) in terms of speed and noise suppression. This is of course a stronger claim because the multivariate (MSE) design is `cheating’.
PS: I forgot to link the GDP-data in my previous entry so here it is: GDP1 and GDP2 . This data is called by the R-code of the MDFA-legacy project.
In the current `state of affairs’ it is impossible to up-load R-code to the Blog. Therefore I had to find another solution. You now have access to the
R-code is ready but I’m unable to upload the files because of security reasons. I’ll find a solution… In the meantime here’s an updated version of the book MDFA-Legacy. You may have a look at the new sections in chapters 2-4. I finally managed to tackle the tedious i1=F,i2=T case in my code.
Here’s a first version of the MDFA-Legacy book with chapters about
- Filter revisions
- Filter Constraints
see MDFA-Legacy. The book is generated in the Sweave environment and I’ll post the R-code for replicating results (tables, graphs) soon.
I’m currently working on the MSE-section of the MDFA book-project. A first draft will be released soon. I make extensive usage of my DFA-manuscript: I urge interested readers to review section 4.1 and, in particular, exercises 1 and 2 (in section 4.1.1). This DFA-material will be `copy-pasted’ and generalized to a multivariate 2-dim setting. Ideas and concepts developed in these exercises will assumed to be known.
In my previous entry I presented a skeleton of the forthcoming MDFA-Legacy project. I here offer insights into the general proceeding: the way I want to instruct and guide interested readers.
I received emails according to which the documents linked in my previous blog-entry could not be downloaded. In the meantime I found a patch: download should now be functional (I will address a `permanent’ stable solution next week).
Back to the topic of this entry… I’d like to offer a teaser: the skeleton of the forthcoming MDFA-Legacy project: MDFA_Legacy (left-clicking should open the document in a new window).
As claimed in my previous entry I wand to collect all spread-out material about MDFA (Multivariate Direct Filter Approach) into a single – monolithic – book format: this forthcoming project is called MDFA-Legacy. As claimed, also, I won’t start from scratch; instead I’ll assume that the material collected in my econometrics-script about DFA (univariate) is known by the interested reader. I here provide a link to this document (left-click to open in new window): DFA. The accompanying (univariate DFA) R-code can be downloaded here (left-click to open in new window): DFA-R-Code.
The somehow dramatic title of this entry might – erroneously – suggest that my days are counted. Well, conditionally on what I know – the full information universe available to me – the likelihood of dying soon is `negligible’ (in the sense of: I don’t spend much time thinking about it). But I received some long-waiting (not to be confounded with `long awaited’) inspiration. This Blog-entry is about a forthcoming project that I’d like to develope on SEFBlog. An older idea which grew-up and became a certitude.
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…