The distribution of values for each attribute is stored in the form of an equi-width histogram. Using this histogram, the estimated size for a non-equality selection when compared to no histogram being used
1. remains the same
2. improves
3. deteriorates
4. can be exactly predicted always
