Advanced Search
All Posts : Product Tips & Techniques

With a small dataset, a summary number can hide more than it reveals. A mean, a median, or a single trend line can look clean while hiding an outlier, a gap, or two distinct groups in the data.

 

Before summarising, plot the individual points. Look at the shape: are the values spread evenly, clustered, or split? Is one value doing most of the work in pulling the average up or down? Only after that visual check does a summary statistic earn its place — and when you do report one, say what the plot showed alongside it, so a reader can see whether the summary is a fair description of the data or just a tidy one.

 

The habit scales down as well as up: it is easy to skip the plot when there are only a dozen values, but that is exactly when a single unusual value has the most influence, and when looking is cheapest.

Please Wait...