Another possibility for showing distributions is to use a histogram. A histogram looks similar to a bar chart, but the bars show a count of occurrences of a value. For example, standardized test auditors looking for evidence of grade tampering might construct a histogram of student test scores. Typically, a distribution might look like this:
The test scores are shown on the x axis and the height of each bar shows the number of students that made that particular score. A typical distribution should have a fairly recognizable bell curve, with some students doing poorly, some doing extremely well, and most falling toward somewhere in the middle.
What if auditors saw something like this?
Something is clearly wrong. Perhaps graders have bumped up students who were just shy of passing to barely passing. It's also possible this may indicate bias in subjective grading instead of blatant tampering. We shouldn't jump to conclusions, but the pattern is not normal and requires investigation. Histograms are very useful in catching anomalies like this.
Let's say you'd like to see a histogram of the time it takes to begin patient treatment. You might start with a blank view and observe steps such as the following:
- Click to select the Minutes to Service field under Measures in the data pane.
- Expand Show Me if necessary and select the histogram.
Upon selecting the histogram, Tableau builds the chart by creating a new dimension, Minutes to Service (bin), which is used in the view, along with a COUNT of Minutes to Service to render the view:
You can see the curve, which peaks at just over 40 minutes and then tapers off with a few patients having to wait as long as 110 minutes. The key to the histogram is the bin.
You can create new bins on your own by right-clicking a numeric field and selecting Create | Bins. You may also edit an existing bin field by right-clicking the bin field itself and selecting Edit. When you first create a bin (or when Tableau creates one based on Show Me), Tableau uses an algorithm to determine a best size for the bin.
In the case of the histogram you created earlier, the size was set at 6.5 minutes. This may not be the most helpful for understanding the data in this case, so you might wish to edit the bin size by right-clicking the Minutes to Service (bin) field under Dimensions in the data pane, selecting Edit, and then adjusting the size to something more intuitive, such as 5:
You'll also want to decide what you want to count for each bin and place that on Rows. When you used Show Me, Tableau placed the COUNT of Minutes to Service on Rows, which is just a count of every record where the value was not null. In this case, that's equivalent to a count of patient visits. However, if you wanted to count the number of unique patients, you might consider replacing the field in the view with COUNTD([Patient ID]).
Just like dates, when the bin field in the view is discrete, the drop-down menu includes an option for Show Missing Values. If you use a discrete bin field, you may wish to use this option to avoid distorting the visualization and to identify what values don't occur in the data.