[pdp-discuss] Take 2 for ClosestEventStat

[rama0042]

Hi All,

A few months ago I e-mailed the list asking about ClosestEventStat.  In
particular, I asked why when using that statistic in LEABRA++, it always
returned the name of the target event in the field for the closest event,
yet always returned a 0 for the distance.

Dr. O'Reilly replied that that stat had to be put in the Trial.loop_stats
not the in final_stats.

I tried the suggestion, and got the same erroneous results.  I let the
matter drop for the time, going on to other things.  I am now at a point
where solving this problem is the only thing left for me to do before I
start running experiments using my network.  So, I tried all kinds of
variations on the ClosestEventStat.  I tried varying where I put it, what
process I associated it with, etc., and I always get erroneous results.  It
either always spits out the same event name as the target (but with a
distance of 0, even though it is apparently always correct), or it always
spits out the name of the first event, for *all* trials, and always gives
the distance as 1 even though it should be 0.  This happens for all trials
except for the one trial where the target actually is the first event, in
which case it gives the distance as 0 even though it should read 1.

I have decided that if no one can shed light on this, then I will just save
the raw data of my experimental runs to a file, then import the data into
MATLAB and write a MATLAB program to do what ClosestEventStat is supposed
to do.

I just thought I would send out one last (plaintive) e-mail to see if
anyone had any ideas.  It would sure be a shame to spend the time writing a
MATLAB program if others have some idea of how to proceed.

Thanks for any help anyone can give.  (And my apologies to the list for
belaboring the question.)

Pradeep

 
Pradeep Ramanathan, M.S., M.A., CCC-SLP
Speech-Language Pathologist/PhD Student
University of Minnesota