Descriptive Statistics
Descriptive statistics tables show summaries for variables with imputed values. A separate table is produced for each variable. The types of statistics shown depend on whether the variable is scale or categorical.
Statistics for scale variables include the count, mean, standard deviation, minimum, and maximum, displayed for the original data, each set of imputed values, and each complete dataset (combining the original data and imputed values).
The descriptive statistics table for tenure (Months with service) shows means and standard deviations in each set of imputed values roughly equal to those in the original data; however, an immediate problem presents itself when you look at the minimum and see that negative values for tenure have been imputed.
For categorical variables, statistics include count and percent by category for the original data, imputed values, and complete data. The table for marital (Marital status) has an interesting result in that, for the imputed values, a greater proportion of the cases are estimated as being married than in the original data. This could be due to random variation; alternatively the chance of being missing may be related to value of this variable.
Like tenure, and all the other scale variables, income (Household income in thousands) shows negative imputed values — clearly, we will need to run a custom model with constraints on certain variables. However, income shows other potential problems. The mean values for each imputation are considerably higher than for the original data, and the maximum values for each imputation are considerably lower than for the original data. The distribution of income tends to be highly right-skew, so this could be the source of the problem.