latqcdtools.statistics.autocorrelation
getTauInt(ts, nbins, tpickMax, acoutfileName='acor.d', showPlot=False)
Given a time series, return estimates for the integrated autocorrelation time and its error.
INPUT:
tpickMax--The largest nt where you think your estimate might become unreliable.
nbins--The number of jackknife bins (for estimating the error in tau_int)
ts--Time series array of measurements. Must be taken from equilibrium ensemble so that
time translation invariance holds. List must be in order of markov chain generation
OUTPUT:
tau_int--Estimate for integrated autocorrelation time.
tau_inte--Its (jackknife) error bar.
itpick--The Monte Carlo separation at which this method found its estimate for tau_int.
remove1Jackknife(ts) -> numpy.ndarray
Create remove-1 jackknife list from 1-d series.
Args:
ts (array-like): time series
Returns:
np.array: 1-d array of jackknife means
tauint(nt, ts, xhat=None) -> numpy.ndarray
Given a time series, calculate estimators for its integrated autocorrelation time at each Markov time separation.
INPUT:
nt--The largest you think tau_int could be.
ts--Time series array of measurments. Must be taken from equilibrium ensemble so that
time translation invariance holds. List must be in order of Markov chain generation.
xhat--True mean of time series (if you know it).
OUTPUT:
acint--List of integrated autocorrelation times.
tauintj(nt, nbins, ts, xhat=None) -> numpy.ndarray
Given a time series, calculate jackknife bins of integrated autocorrelation time for each Markov time separation.
INPUT:
nt--The largest nt at which you think your estimate for tau_int could lie.
nbins--The number of jackknife bins.
ts--Time series array of measurements. Must be taken from equilibrium ensemble so that
time translation invariance holds. List must be in order of markov chain generation
xhat--True mean of time series (if you know it).
OUTPUT:
acintj--2D list indexed by time, then bin number acintj[it][ibin]