[GT] Re: EMA incorrect when using --nb-item

["Robert A. Schmied" <ras AT acm.org>]

Nick Fantes Huege wrote:
> Hello All,
>
> I don't know if this is intentional, but when calculating EMA (and my
> guess is that the same is true for all other indicators that give more
> wight to the more recent dates) the values produced are incorrect if
> --nb-item is used, or if --start and --end are used.
>
> Example:
> ./display_indicator.pl --full --nb-item=10 I:EMA YHOO 60
> ...
> ...
> EMA[60, {I:Prices CLOSE}][2010/07/19] = 15.7129
> 15.712 is wrong value for YHOO's EMA for that day.
>
> After analyzing EMA.pm, it seems the problem is that it does its
> calculations based on the last 10 (or --nb-item) prices.
> A simple fix would be to add $first = 1 inside the
> calculate_interval() subroutine, thus giving it the full range of
> prices to work on.
>
> Nick

aloha nick

the options -full and -nb-items are thought to be mutually exclusive
(but are not processed that way by the code unfortunately).

generally speaking --full, --start=<date>, --end=<date>, --nb-item=<nr>
are used to set the time period to run the analysis on. by default
(without any of these options) only the most recent 200 bars are used,
which implicitly sets --end.

the analysis range can be controlled by --full (run analysis on every bar
in database), or by using one or two of --start, --end or --nb-item.
in this situation, the total number of bars used would be computed from
the fixed reference, either the explicitly set --start or --end
(either the implicit value or the explicitly set one).

i'm not sure why you would want to use --nb-item=10 and also use a 60
bar time period ema?

as far as your analysis of the GT::Indicators::EMA::calculate_interval
method; it is believed that when passed a valid interval range the computed
ema values are correct. this is implicitly suggesting when used as you
describe, display_indicator.pl is not passing a valid interval range.

ras