I am using a real data set that has many zeros (Data point) in it to fit a probability model. The Anderson Darling (AD) Test Statistic for such data gives me infinity which is not correct. I wonder whether we can get AD values for such kind of data or not.