Discussion Of The Imputation Methods

Draft Only: Subjected to Revision due to lack of Reference Citations and Corrections Discussion of the Imputation Methods The Overall Mean Imputation (OMI) is the process by which missing data is imputed by the mean of the results of the first visit partial nonresponse variable. It is the simplest of the all imputation methods because it simply replaces all the missing values by one value. However, this method is not commonly used because it is rather inaccurate especially in cases when the results of the actual value significantly differ from the results of the units that were to missing. A case of this may occur when the study consists of people coming from different age groups and the results vary significantly. There are certain advantage and disadvantage of this method. The advantage of using this method lies in the concept of it being universal. This means that this method can be applied to any continuous set of data. However, it also has numerous disadvantages. Since the missing observations were to be substituted with a single value, its distortion of the distribution of values makes it unsuitable for many other forms of analysis. The Hot Deck Imputation (HDI3) is the process by which the missing observations are imputed by choosing a value from the first visit partial nonresponse variable in its imputation class. This value is either selected at random, or in some deterministic way with or without replacement, or based on a measure of distance. In this study, the missing observations were substituted by choosing a random observation from the first visit partial nonresponse variable. The process of imputing missing values begins with the formation of imputation classes. Imputation classes defined in the methodology were classification groups that were homogeneous. In each imputation class, the set of observations to be substituted and that are substituted are called donors and recipients respectively. Further discussion of the formation of imputation classes will be tackled later on. The Hot Deck procedure of making imputation classes from the total sample is a way of getting the estimates to reflect more accurately the target population. If the matching variables are closely associated with the partial nonresponse variables, then as the imputation classes become smaller and more homogeneous, the nonresponse bias should be reduced. A quality of hot-deck imputation is its ability to produce a clean data set. Actually this quality is present in all imputation methods. The fact that hot deck imputation makes a complete data set out of incomplete data is not to be interpreted as if it was originally complete. If it were treated as complete data set without originally being incomplete then this would be totally misleading in the sense that there is bias in the imputation procedure but it is ignored because it is assumed that it was complete. A method applied to avoid this mistake is by flagging the missing data so that they could keep track of which part of the data set was imputed by using hot deck. Lastly, one of the primary principles that govern the construction of hot deck procedures is to reduce the bias while preserving joint and marginal distributions. This is because hot deck imputation allows more variability in the imputation of missing values compared to the overall mean imputation wherein every missing value is replaced by a constant value, the overall mean of the first visit partial nonresponse variable.