Okay, okay, so measurement and other estimation uncertainties mean that not every detail is significant, and there will always be variation between events. That’s fine, point taken, but what about the big picture? Overall, how do the circulation shifts mentioned earlier on [\ref{fig:walker_elnino}] map onto the region?

This is where the composites finally come in. The idea is to make some type of estimate of the central tendencies of the grouped events, to find some way condense the information and filter out the noise of the individual variability¹¹Not because the individual events are not important, they are after all what we actually experience, but to look for patterns. If I ask you about ENSO, and you hand me seventy-five maps and forty years of crop yield reports, I’ll probably ask if you could start by just giving me a general idea. Of course after being told the general idea, I’ll probably then want to point out that the notion you’ve given me is extremely vague, and doesn’t really describe the situation. At which point you’ll tell me that that’s what I asked for, and to please stop being difficult..

The procedure is to first take all the time slices identified as being in El Niño periods, so those shown previously [\ref{fig:elnino_samples}] plus sixty-three others, and then calculate the arithmetic mean value for each grid cell. Or the median value, or the mode²²Since the anomaly values are continuous, as in they can be different from each other by very small fractions, (\(-1.095\), \(-1.093\), \(-1.101\), etc.), it doesn’t really make sense to talk about the ”most frequent value, unless we ’bin’ the data. That is, take every value between \(-1.25\) and \(-0.75\) and call them all \(-1\), take every value between \(-0.75\) and \(-0.25\) and call them all \(-0.5\), and so on. Obviously this is a fairly crude abstraction of the data, but on the other hand it points most directly to the question ”what is happening most often?”. There are always trade offs.. All of which give subtly different information about the data being analyzed, and so create different maps [\ref{fig:comp_examples}].