Aggregation effects

Where averages hide reversals.

Simpson's paradox

A trend present in every subgroup can reverse when the groups are combined, usually because the groups differ in size and baseline. Both charts are accurate.

Always ask whether the total hides opposing patterns.
Breaking data down by an obvious grouping is the first check.

Choosing bins

Histogram bin width and map colour breaks change the shape and the message. Trying several is honest; presenting only the flattering one is not.