Many sets of data have missing values. If the probability that an observation is missing is independent of its value and of all the variables of interest, then it is said to be missing completely at random (MCAR); an example would be if information on a sampled individual was accidentally lost. If the probability that an observation is missing depends on its value, but not on the values of other observations, then it is described as missing at random (MAR); an example would be the reluctance of individuals earning high incomes to reveal their incomes. Other missing values fall into the category missing not at random (MNAR). MCAR leads to a reduced sample size, but no other problems. Methods for dealing with MAR and MNAR include the EM algorithm and imputation.