I tried an experiment with a basis having a zero bias, e.g.
plot k=. 1>.100<.+/\"1 d=.(?5 200$7) { _1 2 _3 4 _5 6 _3
(zero bias because 0-:+/_1 2 _3 4 _5 6 _3)
and the result still gives the impression of being upward-sloping. However,
I think this is just a consequence of how we tend to ignore the flat line at
the bottom and that the basis numbers are small relative to the capped
range, plus the lengths of the series (200) are relatively small. Thus, the
only lines that cross the middle of the graph (to eventually hit the upper
cap) in this short sample are those that happen to be positively biased.
You can see how a set of longer series like
plot k=. 1>.100<.+/\"1 d=.(?5 2000$7) { _1 2 _3 4 _5 6 _3
has a different appearance.
On 1/22/08, Devon McCormick <[EMAIL PROTECTED]> wrote:
>
> I think what you want to do is something like
>
> plot k=. 1>.100<.+/\"1 d=.(?5 200$6) { _1 2 _3 4 _5 6
>
> (to see 5 lines at a time)
>
> The upward slope is a consequence of this:
>
> +/_1 2 _3 4 _5 6
> 3
>
> i.e. you have an average bias of positive 3.
>
> And this is a consequence not of sequential thinking but maybe of scalar
> thinking.
> FWIW, I was initially puzzled by your result as well: your code looked
> plausible at first glance.
>
> On 1/22/08, Devon McCormick <[EMAIL PROTECTED]> wrote:
> >
> > To clarify with an example:
> >
> > foo/\40 _30 _30 _30 _10
> > 40 10 41 41 41
> >
> > So, lining the input vector up with successive intermediate results
> > shows us the following:
> >
> > 40 _30 _30 _30 _10
> > 1 NB. capped
> > 1 NB. capped
> > 1 NB. capped
> > 41 NB. not capped
> >
> > From this we see that catenating negative numbers to the right will keep
> > feeding a one to be the number added to 40 for the final result. In fact if
> > the sum of the numbers to the right of this adds up to 100 or less (which
> > has to be true because of the upper limit), it will get capped to 1 because
> > the succession of negative numbers will reduce it to that.
> >
> >
> >
> > On 1/22/08, Raul Miller < [EMAIL PROTECTED]> wrote:
> > >
> > > I think you should be using
> > > plot k=:foo/\.d=:(?200$6) { _1 2 _3 4 _5 6
> > >
> > > Remember that 1 foo 2 foo 3 foo 4 is equivalent
> > > to 1 foo (2 foo (3 foo (4))) but that \ works against
> > > prefixes.
> > >
> > > In other words, you are seeing secondary effects
> > > of previous intermediate results having been
> > > clamped when you use foo/\
> > >
> > > --
> > > Raul
> > > ----------------------------------------------------------------------
> > > For information about J forums see http://www.jsoftware.com/forums.htm
> > >
> >
> >
> >
> > --
> > Devon McCormick, CFA
> > ^me^ at acm.
> > org is my
> > preferred e-mail
> >
>
>
>
> --
> Devon McCormick, CFA
> ^me^ at acm.
> org is my
> preferred e-mail
>
--
Devon McCormick, CFA
^me^ at acm.
org is my
preferred e-mail
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
