[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Seasonals and technical analysis



PureBytes Links

Trading Reference Links

Thanks for your emails

Seasonal Adjustment (SA) processes, i.e., X-12, Demetra, SABL etc. are
important in technical analysis because the basic assumption is that time
series data can be broken down into 4 components: seasonal, trend, cyclical,
and random/noise.

Even if you are not concerned with seasonal variations per se, including
"day-of-the-week" seasonals, the other three components are an important
part of analysis. At
the very least you probably subtract "trend" from your system tester results
to obtain a realistic return.

Most of the discussion on the List is about ad-hoc methods and rules patched
together. There is no common methodology or statistical basis. Even the
"leaders" in the field TASC, Futures Mag, Metastock etc. work from a "quasi
empirical" basis without an established method.

The papers produced by the different SA agencies and groups at least have an
open discussion about methodology, i.e., the use of moving averages,
filters, etc.

One caution, you need to be careful in using Seasonal Adjustment processes
because of how
they deal with outliers.

As with most technical analysis, the SA processes
assume and process towards lognormal distributions, i.e., normal
distributions under the curve.

As a trader, this is dangerous thinking because of the reality of
leptokurtic distributions in market time series data. Leptokurtosis implies
much fatter tails where all the action happens and much higher and narrower
central portions where "nothing happens". The economic gain or damage in the
tails can be "very interesting".

Best regards

Walter