Re: A Look At Decimation

[Jeffrey Harteam <jharteam@xxxxxxxxxxxxxx>]

Clyde Lee wrote:

> Decimation is a term that specifies the number of bars that are
> used to make up one bar during analysis.
>
> If we say the decimation rate is two then the bars that are
> analyzed are each composed of the highest high of two adjacent
> bars, the lowest low of the two adjacent bars, the open is
> the open of the first bar and the close is the close of the
> second bar.
>
> This concept can be extended to any number of bars desired.
>
> If we specify 3 as the decimation/compression factor then the
> high of each bar analyzed is the highest high of the current and
> the past two bars, the low is the lowest low of the three bars
> under consideration and the open is the open of the first of the
> three bars and the close is the close of the last bar.
>
> See the following example:
>
> Three             Decimated
> Bars              Bar for
>                   Analysis
>
>            |        |
>       |    |-       |-
>  |    |    |        |
>  |-   |-  -|        |
>  |    |    |        |
> -|   -|            -|
>  |                  |
>
> Now what I do is to roll along and do this to every bar and
> store the "decimated" bars in another array and run the analysis
> on this set of data.
>
> This amounts to a smoothing operation of sorts BUT it quickly
> adjusts when direction changes.
>
> Clyde Lee

 Greetings Clyde:

Thanks for the explaination of this 'rolling operation' of SM.  What I would
like to inquire is in the DOS version of the SM, we get a choice of rolling
the bars from 2-9 bars.  The question is how are we going to decide or
rather, are there any general guideline(s) as to the number of bars that need
to be decimated?  In your experience, what would be the 'more objective'
number of bars that need to be decimated?  Thanks in advance for the answers
and the SM.  Regards

Have a good one
Jeff Harteam
Hong Kong