While computing IQR with out outliers, should we not drop data from both sides as was done with mean? In the example with farm output, only extreme right data gets dropped. Any comments/points?
When we are computing IQR at that we are only calculating with Q2 and Q3 that means Q1 and Q4 is not included.
Thanks, it is true that for computing IQR, we are only concerned with Q2 & Q3 data and IQR is not sensitive to outliers (will not change drastically). But that was not my point, my point was about computation of L25 and L75. Should it be done with 19 elements (as has been done) or with 18, dropping one value in Q1 as well?
It is a good question by @kaybeesan. Any explanation, please?
No, L25 and L75 will be computed on complete dataset only(all 19 elements). As finding IQR every time on trimmed data, it’s possible that we have to drop some more values again if they lie outside the “trimmed data IQR”.