Furthermore, usual normality tests (Shapiro & Wilk, 1965) and
homoscedasticity (Breusch & Pagan, 1979) are required to verify the usual regression assumptions.
The Breusch-Pagan test was not significant: [X.sup.2.sub.(1)] = 0.14, p = 0.15, which expresses the
homoscedasticity of residuals in the regression models calculated.
In general, due to inference reasons, it is assumed that the errors are independent and identically distributed normal random variables with a zero-mean and constant variance [[sigma].sup.2] [I.sub.n], where [I.sub.n] is the identity matrix of n order (
homoscedasticity of the variances).
The between-group comparisons for clinical variables were analyzed by applying the following algorithm: first, each variable was tested for normality or log normality distribution by using the Shapiro-Wilk test and for
homoscedasticity by using the F test.
Assumptions for linearity, independence of residuals,
homoscedasticity, multicollinearity, outliers, and normality were all verified.
The normality and
homoscedasticity of the dependent variables were measured by means of the Shapiro-Wilk and Levene's tests, respectively.
All data was checked for
homoscedasticity, normality, and linearity.
Observations regarding the survival, H, and DBH variables were subjected to the ShapiroWilk normality test, while the
homoscedasticity of the variances was subjected to Anscombe and Tukey's test (1963) at 5% significance level.
The assumptions of normality, sphericity, and
homoscedasticity were met.
In the preliminary analysis, we examined Mahalanobis distance scores ([D.sup.2]), linearity, and
homoscedasticity (Tabachnick & Fidell, 2007).
Partial plots, histogram and normal probability of the residuals were checked to test the
homoscedasticity and linearity of the model, and the independence and normal distribution of the errors.