Influential observations in a dataset are those that have a significant effect on parameter estimation when excluded. In the social sciences, such data points can significantly distort results, leading to interpretations and conclusions that are erroneous. Statistical considerations for excluding influential observations are presented here using an example in economics. This example has two independent variables, but it would be applicable to other analyses having more variables. The techniques considered here are particularly useful when there is more than one independent variable because the influence of one specific data point is not as obvious.
There are no reviews yet.