Briefly |
An Ntigram (pronounced en-ti-gram) is a type of histogram. It shows univariate distributions. |
Usually, histograms have equal-width bins with different counts. In contrast, ntigrams have equal-count bins with different widths.
An advantage of ntigrams is that they show the details of the distribution where there are lots of cases, but wash out the details where there are few cases.
A disadvantage is that the other axis is not "count" as with the traditional histogram, but "density"--which is the number of cases per whatever.
Here are two histograms and an ntigram showing the same data (130 people's ages from Berkeley, California):
A histogram with wide bins. You can clearly see the "hump" in population. | |
A histogram with narrow bins. You can see more detail in the distribution. Look especially at the decline among teenagers, the spike in the early thirties, and the blip at 85. But with such small numbers, do these features really reflect the population? | |
An ntigram with ten bins of 13 people each. The biggest spike is in the early twenties (graduate students at the University) but the other details in low-density areas have been washed out. The vertical axis is Density, in people per year. |
Note: the real, official definition of histograms has density on the vertical axis. What we usually use - equally-spaced bins, "count" on the other axis - is a special case. The point is that population is proportional to area. So an ntigram is really just another kind of histogram - a different special case.