What is the best approach to dealing with missing data? Given that the ideal does not always happen, give two other possibilities, and compare the two.

Question

What will be an ideal response?

Answer 1

Ans: Missing data are a problem for a very clear reason—with missing data, a participant cannot be included in analyses, and power is lost. Since the main reason for conducting research is to determine if there is a difference between our groups, we want as much power as we can get. The best approach to missing data is not to have any. The best way to not have missing data is to have good communication skills and to develop a relationship between participants and researcher. This translates into being available if the participants have questions. If questions are answered, and participants are encouraged to give honest answers, missing data will be at a minimum. Still, missing data happen. There are several ways to deal with this problem. The default for SPSS is pairwise deletion in that the data of a participant with missing data for a particular scale are deleted for that analysis. This is okay if only a few participants are missing data or if you have plenty of participants so that missing a few will not hurt your power that greatly. Most of the time, however, we do not want to lose a participant if we can help it. If the data are continuous and part of a scale, the researcher can use imputation. That is, the researcher can calculate the mean for a particular person for the particular scale and insert that mean into the missing data slot. This can only be done for a limited number of items and has been criticized but has the advantage of being based on individual data. In conclusion, there may be more sophisticated techniques out there, but these are the most likely to be used and are common in the literature.