I am looking at a legacy process for work. We look at corporate budget data. The data is composed of annual budgets for different major categories for different corporate regions. The upper management wants to see medians, and I noticed whoever started this report decided to trim whatever amount below 5k and then time the median of all over budgets. (I think his thought process was that rarely people would have a budget less than 5k , so the data collection/reporting must have been incomplete or incorrect). I have never seen people asymmetrically trim the dataset for medians since I think calculation of median depend heavily on symmetry. I did a quick comparison of trends based on medians without his trimming process and with his trimming process, and there are (i think) significant difference.....what do you guys think? Is this type of trimming for median calculation "appropriate"?