For the pain score data replace each missing value for a subject by the mean of the values the subject actually has for the 30–180 minute observations and then apply the summary measure method to the data using the mean of the six observations that each subject now has. What do you consider to be the problems of this approach?