vlt.neuro.spiketrains.spiketimes_correlation_timeshiftsubtract

  SPIKETIMES_CORRELATION_TIMESHIFTSUBTRACT - Correlation that subtracts correlation not due to precise times

   [XCORR_ACTUAL, XCORR_SHIFTED, XCORR_PERCENTILE, LAGS] = 
     vlt.neuro.spiketrains.spiketimes_correlation_timeshiftsubtract(TRAIN1, TRAIN2, BINSIZE,...
     MAXLAG, SHIFT, TRIALS)

   Computes both the standard spike train correlation (XCORR_ACTUAL) between
   TRAIN1 and TRAIN2, lists of spike times, using vlt.neuro.spiketrains.spiketimes_correlation, and
   computes an estimate of the amount of this correlation that is due to slow
   covarying spike rates with a time scale of SHIFT.

   BINSIZE is the resolution (e.g., 0.002 for 2ms bins), and MAXLAG is the
   maximum lag between the two trains TRAIN1 and TRAIN2 that will be computed.
   (See help vlt.neuro.spiketrains.spiketimes_correlation)

   The estimate of the amount of correlation that is due to slow covarying
   spike rates is computed over TRIALS trials (e.g., 100), during which
   each spike in the trains is time-shifted by a variable amount between 0 
   and SHIFT (uniform distribution). SHIFT might be 0.100 to shift 100ms, 
   for example.  This estimate is returned in XCORR_SHIFTED.  The percentile
   distribution of XCORR_SHIFTED is returned in XCORR_PERCENTILE. Each Nth
   row of XCORR_PERCENTILE is the N-1th percentile of the distribution.