Imputation has become one of the most popular tools used to deal with missing value problems in survey data analyses. This paper compares 11 popular imputation methods. Results indicate "best" performing methods when the population is normally distributed and indicate that the performance of most of these methods deteriorates as the population deviates from the normal distribution. A least trimmed squares (LTS) regression approach is used to implement an automated imputation procedure in SAS that is shown to be robust with outlier data or skewed data. The robust method is also shown to perform well when applied to real world skewed data from a national survey of substance abuse treatment facilities.