Histograms and Bin Width

A histogram splits the range of a numeric variable into equal-width intervals called bins, then draws a bar over each bin whose height shows how many data values fall inside it. Build one by picking a bin width, marking off bins along the axis starting from some chosen value, counting how many data points land in each bin, and drawing bars with no gaps between them. Those touching bars are the whole point: the horizontal axis is a continuous number line, and each bar's width represents a real range of values, not a single category.

Bin width controls what you see. Narrow bins keep fine detail, so a data set with two genuine clusters can show up as clearly bimodal. Widen the bins and nearby values get swept into the same bar, smoothing bumps and gaps away until the same data looks like one smooth mound. Neither picture is wrong. Both come from the same numbers; they just answer the question 'how many groups should I lump these into?' differently. That is why checking a histogram at more than one bin width is a reasonable habit before deciding on a shape to report.

This sits right next to describing distributions by shape, center, and spread. A histogram is where you first read off whether a distribution looks roughly symmetric or skewed, where the center and typical spread are, and whether there are outliers, gaps, or multiple modes. Bin width is the dial that can create or hide that last set of features, so when you report shape you're really reporting what one particular bin width happened to reveal.