Next we tested for homogeneity of variance in asymmetry scores between sexes using Levene's test on each scute pair, to ensure that the assumptions of homoscedasticity
were not violated.
In the final (fourth) step, we check for the potential problem of homoscedasticity
as all models are estimated based on cross-sectional data.
The regression models were evaluated for compliance with the linearity, normality, homoscedasticity
and independent error terms assumptions of multiple regression (Hair, Anderson, Tatham and Black 1998), and no violations were observed.
In testing for assumptions of normality and homoscedasticity
, Kolmogorov-Smirov tests were used and distributions were explored using various graphical techniques (Coakes & Steed 1999).
The assumptions of normality and homoscedasticity
are tested in all regressions that follow.
Firstly, conditional homoscedasticity
(5) for the residuals are explicitly modeled.
Using the White's test, the homoscedasticity
assumption of the error process is supported.
Original data did not fulfill requirements of normality and homoscedasticity
A preliminary examination for outliers, excessive multicollinearity, and departures from linearity, homoscedasticity
, and normality was performed.
prostrata failed the assumptions for normality and homoscedasticity
and a Wilcoxon-Signed Rank test was used (NCSS 1995)).
Second, we employed the Goldfeld-Quandt test for homoscedasticity
(Goldfeld & Quandt, 1965) to assess the degree of heteroscedasticity in the distribution of error terms.
Data were log transformed as needed to satisfy the assumption of homoscedasticity