12.1 Error in Linear Regression Models

straight enough condition

the scatterplot of y against x is roughly linear - this is part of the linearity assumption

outlier condition

the distribution of data should not have outliers - this is part of the Normal errors assumption

independence assumption

the errors in the underlying regression model are independent of each other

error term

the difference between what you expect to see and what you observe

does the plot thicken? condition

the spread around the regression line should be nearly constant - this is part of the equal variance assumption, and it can be checked using a scatterplot

randomization condition

the individuals in the sample are randomly selected - this is part of the independence assumption

random residuals condition

there should be no evidence of patterns, trends, or clumping when looking at the residuals plot - this is part of the independence assumption

nearly Normal condition

the data in the distribution should resemble a Normal model - this is part of the Normal errors assumption, and it can be checked using a histogram of the residuals