I’m sorry if this question had been asked before. How to determine the number of intervals when we’re dealing with a dataset, where we have an observation that is too far away bigger than the mean?
Refer. 1st Dataset of exercice related to skewness. the width of interval is 27.4; the lower observation is 48 and the upper one is 870.
Hi Gilles,
Thanks for reaching out.
In general we are looking for a representation that has between 5 and 20 intervals.
To determine the skewness, I usually start with a histogram with many bins (e.g. 100-1000, depending on the size of the dataset). This gives a good feel about the frequency distribution of the dataset.
Afterwards, I’d normally get the bins down to 20 and see if the general trend remains.
Overall, I try to preserve the general trend while using as few bins as possible.
Best,
Iliya