What are some things that a bar chart with error can tell us?

Question

In the context of this exercise, what are some things that a bar chart with error can tell us?

Answer

Bar charts with error can tell us a few different qualities about the data.

One thing that error bars can tell us is the range of the values. The height of the bars of a bar chart typically represents the average of the values, while the error bars can show us the range of the lowest to the highest value contained in the data. By showing the range of lowest to the highest value, we can know that all the data falls within the error bars.

In addition, the actual sizes of the error bars can tell us how close, or how accurate, the data is. If the error bars are short, then this means that the values are close together. However, if the error bars are large, then the values must be spread out.

3 Likes

Hi, I am confused. In statistics, the error bar stands typically for the standard deviation or confidence interval. It should not be used to show the highest and lowest value. Take this as an example. As we can see, the top of the bar is always in the middle of the error bar. Let’s say we have three numbers, (1,2,5). The average number is 4. So the top of the bar stands for 4. However, the biggest number of this series is 5, only 1 up from 4. Yet, the smallest number is 1, 3 lower from the 4. So the average value, 4 is no way between the error bar if the error bar tell us the highest to the lowest number,

3 Likes

In many cases error bars represent a standard deviation, or a particular confidence interval. However, this is not always the case. It seems that there are some cases where error bars represent maximum and minimum. I think it’s not wrong to say that error bars can represent the range of lowest to the highest value.

Since there are different usages, each chart needs explanation. In any case, error bars don’t make sense unless there is a clear description of what they represents.

By the way, for the example of the three numbers (1,2,5), I think it will look like the leftmost bar in the chart below. The error bar shows the range between 1 and 5.
bar_example

import matplotlib.pyplot as plt

averages = [4, 3, 3, 5]
lowerr = [3, 1, 1, 4]
higherr = [1, 1, 2, 1]

plt.bar(range(4), averages, yerr=[lowerr, higherr], capsize=5)
plt.show()

In the current version of matplotlib (I used 3.2.1), it seems we can specify different -/+ values ​​for yerr.