China
Africa
Explore chapters and articles related to this topic
Hypothesis Testing: Univariate Parametric Tests
Published in Shayne C. Gad, Carrol S. Weil, Statistics and Experimental Design for Toxicologists, 1988
Analysis of covariance (ANCOVA) is a method for comparing sets of data which consist of two variables (treatment and effect, with our effect variable being called the “variate”), when a third variable (called the “covariate”) exists which can be measured but not controlled and which has a definite effect on the variable of interest. In other words, it provides an indirect type of statistical control, allowing us to increase the precision of a study and to remove a potential source of bias. One common example of this is in the analysis of organ weights in toxicity studies. Our true interest here is the effect of our dose or exposure level on the specific organ weights, but most organ weights also increase (in the young, growing animals most commonly used in such studies) in proportion to increases in animal body weight. As we are not here interested in the effect of this covariate (body weight), we measure it to allow adjustment of the measurement of the variate in which we are interested (the organ weights). Analysis of covariance allows us to make this adjustment. We must be careful before using ANCOVA, however, to ensure that the underlying nature of the correspondence between the variate and covariate is such that we can rely on it as a tool for adjustments (Anderson et al.,1980 and Kotz and Johnson, 1982).
Generalized linear models
Published in Kenneth G. Russell, Design of Experiments for Generalized Linear Models, 2018
Earlier examples described situations where the explanatory variables in a linear model were all real variables (Example 1.1.1) or were all indicator variables (Example 1.1.2). However, in some experimental situations, the explanatory variables in a linear model would include both categorical or ordinal variables and also “real” variables. The latter variables were known as covariates, and the analysis of such linear models was called analysis of covariance (ANCOVA).
Wetland Effects on Hydrological and Water Quality Characteristics of a Mid-Michigan River System
Published in Carl C. Trettin, Martin F. Jurgensen, David F. Grigal, Margaret R. Gale, John K. Jeglum, Northern Forested Wetlands, 2018
T. M. Tompkins, W. W. Whipps, L. J. Manor, M. J. Wiley, C. W. Radcliffe, D. M. Majewski
Data were examined statistically using standard ANOVA and ANCOVA techniques. A posteriori contrasts for land use categories were evaluated using Tukey’s HSD test (Wilkenson, 1990). Data distributions for hydrological and chemical parameters were summarized for various land use categories using box plots (McGill et al., 1978).
Estimation of the static bending modulus of elasticity in glulam elements by ultrasound and modal-updating NDT techniques
Published in Mechanics Based Design of Structures and Machines, 2023
Ramon Sancibrian, Ignacio Lombillo, Rebeca Sanchez, Alvaro Gaute-Alonso
The statistical differences between variables were assessed using the coefficient of determination R2. Hypothesis testing was used to verify the independence of the means when comparing results from the same type of test. The analysis of the variance (ANOVA) test was conducted to verify the existence of the regression. The null hypothesis means that the predictors do not provide any information about the response, while the alternative hypothesis checks the existence of the linear relationship between predictors and the response. The analysis of the covariance (ANCOVA) was used to compare two regression lines to each other. This analysis compares the slopes and the intercepts and determine whether they belong to the same or different groups. ANCOVA makes the assumptions of linear regression with normality and homoscedasticity. Significant differences were accepted when p < 0.05.
A simplified skid resistance predicting model for a freeway network to be used in a pavement management system
Published in International Journal of Pavement Engineering, 2022
Heriberto Pérez-Acebo, Mikel Montes-Redondo, Andreas Appelt, Daniel J. Findley
Multiple linear regression (MLR) analysis is a usual statistical technique to analyze the relationship between a quantitative dependent variable (the predicted variable) and various quantitative independent variables (the predicting variables). All the variables must be quantitative; however, it is possible to introduce qualitative independent variables by means of friction variables. Furthermore, Analysis of Variance (ANOVA) is another usual statistical technique that analyzes the relationship between a quantitative dependent variable and various qualitative independent variables (factors). Additionally, The Analysis of Covariance (ANCOVA) is a statistical technique that analyzes the relationship between a quantitative independent variable and both types of independent variables, i.e. quantitative and qualitative independent variables. Finally, the General Linear Multiple (GLM) regression model may be regarded as the most general form of linear regression modelling, because it includes MLR models with quantitative variables and MLR models with qualitative and quantitative variables at the same time, and, therefore, it includes both the ANOVA and ANCOVA models simultaneously.
Aggregate gradation effect on the fatigue performance of recycled asphalt mixtures
Published in Road Materials and Pavement Design, 2021
G. Bharath, K. Sudhakar Reddy, Vivek Tandon, M. Amaranatha Reddy
Analysis of Covariance (ANCOVA) was done to determine the factors having a significant effect on the fatigue resistance of bituminous mixtures. ANCOVA technique combines ANOVA and regression analysis. A significance level (α) of 0.05 was considered to evaluate the significance of the differences in the fatigue performance of different mixes. In this analysis, the significance of the effect of a particular independent variable on the fatigue life (the dependent variable) was evaluated with the effect of the secondary continuous variables (covariates) being statistically controlled. Table 4 gives the “p” values obtained for different cases. The effect of RAP content was evaluated with fatigue life as dependent variable and RAP content and initial tensile strain as independent variable and covariate respectively. The effect of aggregate gradation was assessed for two different RAP contents; 25% and 35% with fatigue life as dependent variable and aggregate gradation and strain level as independent variable and covariate respectively. The p-values obtained from the analysis of the effect of the variation of fatigue life with RAP content are presented in Table 5. P-values for the relative effect of different mixes (gradations) are given in Table 6.