In linear regression, heteroskedasticity occurs when the variance of the residuals (error terms) is not constant across all observations in the sample. If the variance of the residuals remains constant, it is called homoskedasticity, which is one of the classic assumptions of ordinary least squares (OLS) regression. Leptokurtic and platykurtic refer to the kurtosis (tailedness) of a distribution, and multicollinearity refers to high correlations among independent variables.