Algorithm Bias: A Statistical Review: The Biggest Challenge Faced by Technical Analysts Is How Best to Use Domain Knowledge to Infer an Appropriate Bias in Their Algorithms. Here, We Dive into the Subject of Model Selection Using a System for the Pound/dollar

By Sewell, Martin | Futures (Cedar Falls, IA), September 2009 | Go to article overview

Algorithm Bias: A Statistical Review: The Biggest Challenge Faced by Technical Analysts Is How Best to Use Domain Knowledge to Infer an Appropriate Bias in Their Algorithms. Here, We Dive into the Subject of Model Selection Using a System for the Pound/dollar

Sewell, Martin, Futures (Cedar Falls, IA)

The most general (and toughest) challenge faced by technical analysts is neither optimization (optimizing parameters is straightforward) nor overfitting (overfitting avoidance is an assumption), but how best to use domain knowledge to infer an appropriate bias in their algorithms. At the risk of oversimplifying, statistics generally concerns testing a given hypothesis, while machine learning concerns formulating the process of generalization as a search through possible hypotheses in an attempt to find the best hypothesis. Classical statistics involves calculating the probability of the data if the null hypothesis is true, while Bayesian inference involves calculating the probability of a hypothesis, given the data.

As traders in particular, and scientists in general, our aims are better aligned with the paradigms of Bayesian inference and machine learning than with classical statistics.

Consider this: Let B be background information, H a hypothesis and D data. Then P(H|B) is known as the prior, P(D|B) the probability of the data, P(D|HB) the likelihood, and P(H|DB) the posterior. The probabilities are famously related via Bayes' theorem,

P(H|DB) = P(H|B)P(D|HB)/P(D|B)

As there is no such thing as an absolute probability, for notational convenience we often omit B. As the denominator in Bayes' theorem, P(D|B), is independent of H, when comparing hypotheses we can omit it and use:

P(H|D) [infinity] P(H)P(D|H)

In the 18th century, Hume (1740) pointed out that "even after the observation of the frequent or constant conjunction of objects, we have no reason to draw any inference concerning any object beyond those of which we have had experience." More recently, and with increasing rigor, Mitchell (1980), Schaffer (1994) and Wolpert (1996) showed that bias-free learning is futile.

The important point is that one can never generalize beyond known data without making at least some assumptions. The no-free-lunch theorem for supervised machine learning (Wolpert 1996) states that in terms of off-training set error, there are no prior distinctions between learning algorithms. In particular, this implies that there is no free lunch for overfitting avoidance. In other words, we should only constrain our algorithm if this reflects our prior beliefs.

A model is a family of functions, or equivalently, a function is a particular parameter choice of a model. Model selection is the task of choosing a model with the correct inductive bias, which in practice means selecting a model of optimal complexity for the given data. A more complex model will always fit the training data better, but may not represent the true underlying model and thus perform poorly on new data. Note that model selection (which is difficult) logically precedes parameter selection (which is well understood).

OVERFITTING SOLUTION

Below, I present a pedagogical example of Bayesian model selection, a method which in principle solves the overfitting problem and is originally due to the work of Sir Harold Jeffreys 70 years ago (Jeffreys 1939). The aim is to predict the daily British pound/U.S. dollar interbank rate. The data set spans Jan. 1, 1993, to Feb. 3, 2008, and consists of the average ask price for each day. Where the target (defined below) was zero return (weekends), the data were excluded. The training set consisted of 3,402 data points and the out-of-sample set included 1,701 data points.

For reasons of market efficiency, it is safest to take the view that there are no privileged features in financial time series, over and above keeping the inputs potentially relevant, orthogonal and using Tobler's first law of geography that states that "everything is related to everything else, but near things are more related than distant things" (Tobler 1970).

Let p-n be the exchange rate n days ago. Consider five potential inputs:

[x.sub.1] = log([p. ā¦

• Questia's entire collection
• Automatic bibliography creation
• More helpful research tools like notes, citations, and highlights

If you are trying to select text to create highlights or citations, remember that you must now click or tap on the first word, and then click or tap on the last word.
One moment ...
Project items

Highlights (0)
Some of your highlights are legacy items.
Citations (0)
Some of your citations are legacy items.
Notes (0)
Bookmarks (0)

Project items include:
• Saved book/article
• Highlights
• Quotes/citations
• Notes
• Bookmarks
Notes

Cited article

Style
Citations are available only to our active members.
Sign up now to cite pages or passages in MLA, APA and Chicago citation styles.

(Einhorn, 1992, p. 25)

(Einhorn 25)

1

1. Lois J. Einhorn, Abraham Lincoln, the Orator: Penetrating the Lincoln Legend (Westport, CT: Greenwood Press, 1992), 25, http://www.questia.com/read/27419298.

Cited article

Algorithm Bias: A Statistical Review: The Biggest Challenge Faced by Technical Analysts Is How Best to Use Domain Knowledge to Infer an Appropriate Bias in Their Algorithms. Here, We Dive into the Subject of Model Selection Using a System for the Pound/dollar
Settings

Typeface
Text size
Search within

Look up

Look up a word

• Dictionary
• Thesaurus
Please submit a word or phrase above.

Why can't I print more than one page at a time?

Full screen

Cited passage

Style
Citations are available only to our active members.
Sign up now to cite pages or passages in MLA, APA and Chicago citation styles.

"Portraying himself as an honest, ordinary person helped Lincoln identify with his audiences." (Einhorn, 1992, p. 25).

"Portraying himself as an honest, ordinary person helped Lincoln identify with his audiences." (Einhorn 25)

"Portraying himself as an honest, ordinary person helped Lincoln identify with his audiences."1

1. Lois J. Einhorn, Abraham Lincoln, the Orator: Penetrating the Lincoln Legend (Westport, CT: Greenwood Press, 1992), 25, http://www.questia.com/read/27419298.

Welcome to the new Questia Reader

The Questia Reader has been updated to provide you with an even better online reading experience. Ā It is now 100% Responsive, which means you can read our books and articles on any sized device you wish. Ā All of your favorite tools like notes, highlights, and citations are still here, but the way you select text has been updated to be easier to use, especially on touchscreen devices. Ā Here's how:

1. Click or tap the first word you want to select.
2. Click or tap the last word you want to select.

Thanks for trying Questia!

Please continue trying out our research tools, but please note, full functionality is available only to our active members.

Your work will be lost once you leave this Web page.