China
Africa
Explore chapters and articles related to this topic
Classification, Comparison, and Correlation
Published in F. Brent Neal, John C. Russ, Measuring Shape, 2017
Points with a large Cook’s distance, often suggested as greater than 1.0, are considered to merit closer examination in the analysis. A more specific test depends upon the fact that the values obey the F-distribution, and so the distance can be compared to the median value of the F-statistic, which corresponds to a 50% confidence, or placing the point at the edge of the 50% confidence region of the data set. Figure 6.62 shows these median values for various combinations of f and n. The recommended test value of 1.0 is usually a safe, conservative estimate.
Linear Regression
Published in Simon Washington, Matthew Karlaftis, Fred Mannering, Panagiotis Anastasopoulos, Statistical and Econometric Methods for Transportation Data Analysis, 2020
Simon Washington, Matthew Karlaftis, Fred Mannering, Panagiotis Anastasopoulos
Perhaps the most common measure used to assess the influence of an observation is Cook’s distance, D. Cook’s distance quantifies the impact of removal of each observation from the fitted regression function on estimated parameters in the regression function. If the effect is large, then Di is also large. Thus, Cook’s distance provides a relative measure of influence for observations in the regression.
Design of Experiment
Published in Giorgio Luciano, Statistical and Multivariate Analysis in Material Science, 2021
The Scale-Location plot should be constant in an ideal case, since it is useful for checking variance homogeneity. Again, measurements 1, 6, and 7 are marked as the outliers. As a rule of thumb, we should check samples that show values of 1 of Cook’s distance, and deem them suspicious.
Driving collaborative supply risk mitigation in buyer-supplier relationships
Published in Supply Chain Forum: An International Journal, 2021
Deodat Mwesiumo, Bella B. Nujen, Arnt Buvik
The presence of influential observations was checked by assessing Cook’s distance (Cook, 1977). Three cases turned out to be outliers that could bias our analysis, and were thus excluded from the dataset. The final sample of 142 observations is sufficient according to the recommended PLS-SEM sample size for a statistical power of 80% (Hair et al. 2017). Next, we applied graphical and statistical tests to check the normality of the residuals and heteroscedasticity. As shown in Figure 2, most of the observations lie on the line, suggesting that the normality assumption is fulfilled. Likewise, the charts of residuals versus fitted values and standardised residuals versus fitted values show that the residuals are spread almost equally along with the ranges of predictors, suggesting that heteroscedasticity does not exist. We then deployed the ‘gvlma’ package (Peña and Slate 2006) to confirm the assumptions through statistical tests. Table 2 shows that all conditions, including normality (checked through skewness and kurtosis) and absence of heteroscedasticity, are met.
Role of Internet Self-Efficacy and Interactions on Blended Learning Effectiveness
Published in Journal of Computer Information Systems, 2022
Ritanjali Panigrahi, Praveen Ranjan Srivastava, Prabin Kumar Panigrahi, Yogesh K Dwivedi
Before proceeding with the structural model, the multivariate assumptions are tested for influencers and multicollinearity. The influencer analysis is performed in SPSS with Cook’s distance.94 Cook’s distance of less than 0.07 is obtained for all observations which are below the threshold value of 1. The presence of multicollinearity in the dataset is checked in SPSS for identifying the presence of collinearity among predictor variables. The multicollinearity is checked with the Variance Inflation Factor (VIF) and tolerance values. VIF values of less than 3.2 and tolerance values of more than 0.3 are obtained for the constructs (with a required threshold of less than 10 for VIFs and greater than 0.1 for tolerance).90
Direct player observation is needed to accurately quantify heading frequency in youth soccer
Published in Research in Sports Medicine, 2018
Alexandra Harriss, David M. Walton, James P. Dickey
Interrater reliability between the researcher and trained soccer expert was assessed using Cohen’s kappa (κ) (Hallgren, 2012; McHugh, 2012). The Cook’s distance identified the influence of each data point on the regression; values greater than 0.33 (4/n: in this case 4/12) were considered outliers. Based on this criterion, one player was identified as an outlier and removed from analysis. This player participated in 19 of the 20 season games. This analysis was completed using SPSS statistical software (Version 24, IBM Corp, New York, NY, USA).