Moving Frame Tests by M. J. Hinich ********************************************************* US$/Candadian$ Column 1 10 ( 11.90%) significant H frames in run 1 12 ( 14.29%) significant T4 frames Statistics are computed from an AR( 5) fit to each frame Frames are not overlapped Frame Length 90 No. of frames = 84 First datum read in file 1 Last datum read 7569 Sample size = 7560 Sampling interval = 1.000000 day False alarm threshold = 0.05000 C & H Exponent = 0.40 T Exponent = 0.30 C & H Lags = 6 T4 Lags = 4 Bicovariances used = 15 Tricoviances used = 4 100 Bootstrapped Sizes: H = 0.807E-02 C = 0.846E-01 T4 = 0.226E-01 Bootstrapped Standard Errors For Relative Standard Deviations = 0.1145 For Relative Ranges = 0.1129 For Relative Kurtoses = 0.6664 Format: (f10.5,2x,a12) ____________________________________________________________ Descriptive Statistics of Data ============================================================ Mean = 0.558E-04 Std Dev = 0.270E-02 Skew = 0.932E-01 Kurtosis = 3.63 C(6) = 52.05 Max value = 0.191E-01 Min value = -0.184E-01 ************************************************************ Mean One Step Forecast Absolute Error for Significant Frames = 0.0339 Mean One Step Forecast Absolute Error for Insignificant Frames = 0.1442 Standard Deviations: Significant frames = 0.0002 Insignificant = 0.0006 Correlations for 10 largest H Statistics ========================================== Correlation (H,C) = -0.36 Correlation (H,Sig) = -0.06 Correlation (H,R*R) = -0.61 Correlation (C,R*R) = 0.32 Correlation (H,Kurt) = -0.71 Correlation (H,Range) = -0.41 Correlations of Statistics for the Rest of the Windows ====================================================== Correlation (H,C) = -0.12 Correlation (H,Sig) = 0.03 Correlation (H,R*R) = -0.17 Correlation (C,R*R) = -0.08 Correlation (H,Kurt) = 0.31 Correlation (H,Range) = 0.05 Sig-CI are standard deviations that are outside a 99% bootstrapped confidence interval of its bootstrapped mean Ran-CI are ranges that are outside a 99% bootstrapped confidence interval of its bootstrapped mean The mean (Mean), standard dev (Sig), kurtosis (Kurt) and range for each frame are divided by their values computed from all the data to normalize these statistics for plotting Mshift is the Farley-Hinich test of a mean shift Vshift is the Farley-Hinich test of a variance shift The values are cdf probabilities under the null of no shift Version 7-21-2007 Date & Time Fri Mar 21 14:07:10 2008 Elapsed time = 0.02 secs