The purpose of this study was to impute missing values for strangers and other victim/offender relationships using an expectation-maximization algorithm and homicide data. Missing data refers to either unit missing or missing values or both. Unit missing data occur when alternate sources indicate that not all instances of the phenomena have been recorded. Values are characteristics describing objects, and variables are logical grouping of values. Data can be missing completely at random (MCAR), missing at random (MAR), or non-ignorable. The present analyses take advantage of the superior quality of homicide data available from the Los Angeles and Chicago police. It improves upon past research with respect to the categorization of victim/offender relationships. It also seeks to examine whether and how imputations are affected by the set of predictors from which the parameters are estimated. The predictor variables in the two data sets were coded to be as similar as possible. The results confirm that unknown victim/offender relationships are not composed primarily of homicides involving strangers. With clearances as a predictor variable, stranger homicides increased by 5.5 percent. The results suggest that the two existing diametrically opposed claims about missing values are both overdrawn. The findings do not support the argument that missing data resulting from offenders not being arrested makes very little difference. They also do not support the view that there are substantial numbers of stranger homicides represented by missing values. It appears the assignment of cases with unknown victim/offender relationships to known categories on the basis of missing value imputation is influenced by the types of variables available to be used as predictors, with the availability of a clearance status variable being particularly important. 8 tables, 9 footnotes, 57 references