Hi Martin
By removing references to fields and
taking out of the loop some stuff that could be done outside, I got that
down to a little over 4 ticks. The stack isn't cleanly written, but I'll
send the revised version to you. It will be compacted, then Binhexed.
You could just mail me a copy of the revised script. I already have
changed the stack relatively considerably; You are right; I count my
tracking loop at 9+ ticks per frame; speed was not my goal -- as long
as it looks smooth it's OK with me; but I like elegant programs so feel
free to send me the faster version.
In playing with it, I have noticed a consistent deviation between my tracking
behaviour and that of the model. When the mouse track is strongly elliptic
(small determinant), there is an oscillation near the ends of the ellipse,
which shows up as a dent on the side of the end of the ellipse
Yes. I think I get about the same thing. This might be the result of higher
level systems anticipating the change in the disturbance. I think we really
have to do this with a nice, random disturbance instead of the sine wave.
But I had a thought about timing, given my failure with Jon Pugh. The times
are slow enough that ticks can be a reasonable measure.
That's what I thought.
How about making
a long sine table, say 1000 or 5000 elements rather than 200.
I would like to have a long random table. There are various ways to
do this. I might just add some sine waves together (with random
phase) or filter a random number stream. But that was going to be
my next goal -- we've got to eliminate systematic influences from
higher level systems that might be involved to to the predictable
nature of the sine (eliptical) disturbance. Using a random disturbance
should improve the fit of the basic model.
The machine speed would affect the jerkiness
of the presentation, and perhaps not that very much.
What do you think? Should I try it?
Sure. I'll try it too.
I'd like to get on to more complex tracking questions.
Great -- but I want to make sure that we are doing the best we can
with the basic model (especially in the high conflict cases) before
starting to guess about what might be going on that results in discrepencies
from the model.
Subjectively, I
find myself doing far too much planning in following these elliptical
disturbances
Yep; that's why we need the random disturbance; and a long one at that,
so that the subject cannot learn it. That would be my next step. I'm
sure that there will still be systematic discrepencies from the simple,
single level model, even with the random disturbance. But with a
random disturbance I would be more confident that any discrepencies
are "real"; probably the result of reorganization or attention processes.
Of course, the influence of higher level systems with the "eliptical"
disturbance is a "real"phenomenon; but I think we can get a better
handle on that too if we compare the results with the elipses to the
results we get with the random disturbance (when these higher level
"predictive" systems are presumably are not involved).
So send me the script if you like -- I'll work on the new disturbance
this evening (maybe).
Best
Rick
PS. I'm sorry that the psychophysics post was unclear; probably the result
of creating posts in real time (along with a large dollop of juvenile
dementia).
But I think it is an excellent topic so please feel free to make me clarify
any talk that you think is passing through my
hat.