I had a great data crunching / writing day yesterday. This is for my metrology paper on early aes grave (RRC 14 and 18). Then I got off the charts and bar graphs with trendlines and onto histograms. Excel has a lovely function (buried deep in the bowels of the programming) that lets you change the number of bins. This changes the shape of the data. All are true, but all also give a different impression.
Here are the weights of RRC 14/1 as reported by Haeberlin 1910 in both a 20-bin histogram and also a 10-bin histogram. (I obsessively tried out each number of bins from 4 to 25, but I won’t put them all up–its a little obscene. It would be so cool if one could create a little video of these shifting pictures.) The 10-bin histogram shows the strong tendency for weights to be in this 304-336g range, but the 20-bin helps us see better that steep drop off after 344g and the difference in the data shape between too light specimens and too heavy specimens.
I cannot actually put multiple histograms for each type into this article but I do want to communicate the way the histogram is just helping us see the shape of the data, not necessarily a static picture. I sort of feel I need to know more about exploratory statistics, but I also want to get better at drawing pictures and communicating what we can see in the numbers. The numbers themselves often put people off, as do statistical concepts/formulae/jargon.
I’m writing here in hopes it might dawn on me as I write which picture or pictures are most important for this article. Inspiration has not struck. I’m going to keep throwing in charts and cut later. We’ll see where I end up.
(Sorry the charts are boring gray, but they will be cheaper to publish that way, even if poor for the blog.)