High leverage: The data point must be far away from the mean of the predictor variable, where "far away" generally means greater than two standard deviations away from mean in either direction (positive or negative). In other words, the z score of the predictor value should be greater than +2 or less than -2. Every regression line, by definition, passes through the mean of both the predictor and outcome variables. For this reason, you can think of the regression line as a seesaw, pinned to the middle of the scatterplot, pivoting clockwise or counterclockwise around the middle when you remove a data point. The farther away from the middle a data point is, the bigger the potential swing of the regression line. This is just like how sitting on the very end of a seesaw gives you more leverage than sitting close to the middle. Removing a data point near the mean of the predictor variable cannot affect the slope of the regression line very much; removing a data point far away from the mean of the predictor variable may affect the slope quite a lot, but only if the other two criteria are also true. Large residual: The data point must be far away from the regression line, where "far away" generally means greater than two standard deviations away from the line in either direction (above or below). In other words, the standardized residual should be greater than +2 or less than -2. Removing a data point near the regression line cannot affect the slope of the regression line very much