13.07.2015 Views

Consciousness-Based Education - Maharishi University of ...

Consciousness-Based Education - Maharishi University of ...

Consciousness-Based Education - Maharishi University of ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

U.S.-Soviet Relations and the <strong>Maharishi</strong> Effect21The SR statistic (SR = 5.782 for these residuals), is defined as thesample range divided by the standard deviation (residual sum <strong>of</strong> squaresdivided by n = 1). A two-tailed test was used, based on tabled values<strong>of</strong> the SR statistic. The interpolated upper tail .95 critical value for n =93 was 5.8405.22Skewness was –.541 (standard error 0.250) and excess kurtosis 0.998(standard error 0.495).23As reported above, sensitivity analysis also found very similar resultsto those reported in Table 2 for an outlier-adjusted model that hadresiduals for which normality could not be rejected using either the SRor Jarque-Bera tests.24The AR(1) parameter estimate was 0.501 (t = 5.60, p = .000), and theAR(5) estimate was 0.185 (t = 2.05, p = .040). The residual standarderror was 29.070 and R 2 = .378.25The LBQ statistic was insignificant for lags 1–12 (Χ 2 (10) = 7.0, p =.725) and for lags 1–24, (Χ 2 (22) = 13.5, p = .918).26Three negative outliers were detected: April 1981 (IO, t = –3.30),September 1983 (IO, t = –3.76), and February 1984 (TC, t = –4.45).One positive IO outlier was detected for November 1985 (t = 4.89).27The minimum AIC noise model was AR(1), stationary and invertible.The AR(1) estimate was 0.537, t = 5.83, p < .0001. There wereno significant autocorrelations at lags 1–24 and the LBQ statistic wasnot significant for lags 1–12 (Χ 2 (11) = 5.6, p = .899) and for lags 1–24(Χ 2 (23) = 16.4, p = .838). The residual standard error was 22.879. Theplot <strong>of</strong> residuals was consistent with a white noise series with stablevariance and no large outliers.28For the unadjusted model, the only significant correlation betweenparameters was –.26. For the outlier-adjusted estimates, the maximumcorrelation between estimates was –.17, with none being statisticallysignificant.29The correlations between parameter estimates were nearly all 0.30 orlower in absolute value, while the highest correlation (–.47) was thatbetween the constant term and the Gorbachev intervention parameter.30For lags 1–12 the Ljung-Box Q statistic was Χ 2 (8) = 3.1, p = .928, andfor lags 1–24, Χ 2 (20) = 7.2, p = .996.31This was confirmed by senstivity analysis. After outlier detection andadjustment using the default values <strong>of</strong> the SCA oestim procedure, the489

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!