We have already seen that data can be categorised as either numerical or categorical. While bar charts are preferred for displaying categorical data, histograms are the preferred option for data that is numerical.
Continuous numerical data, such as times, heights, weights or temperatures, are based on measurements, so any data value is possible within a large range of values. For displaying this type of data, a special chart called a histogram is used.
As an example, the following frequency distribution table and histogram represent the times taken for $72$72 runners to complete a ten kilometre race.
Class interval | Frequency |
---|---|
$45\le\text{time }<50$45≤time <50 | $9$9 |
$50\le\text{time }<55$50≤time <55 | $7$7 |
$55\le\text{time }<60$55≤time <60 | $20$20 |
$60\le\text{time }<65$60≤time <65 | $30$30 |
$65\le\text{time }<70$65≤time <70 | $6$6 |
The histogram represents the distribution of the data. It allows us to see clearly where all of the recorded times fall along a continuous scale.
What may surprise us at first is that the histogram above has only five columns, even though it represents $72$72 different data values.
To produce the histogram, the data is first grouped into class intervals (also known as classes or bins), using the frequency distribution table.
In the table above,
Every data value must go into exactly one and only one class interval.
Class intervals should be equal width.
There are several different ways that class intervals are defined. Here are some examples with two adjacent class intervals:
Class interval formats | Description | |
---|---|---|
$45<\text{time }\le50$45<time ≤50 | $50<\text{time }\le55$50<time ≤55 |
Upper endpoint included, lower endpoint excluded. |
$45\le\text{time }<50$45≤time <50 | $50\le\text{time }<55$50≤time <55 |
Lower endpoint included, upper endpoint excluded. |
$45$45 to $<50$<50 | $50$50 to $<55$<55 |
Lower endpoint included, upper endpoint excluded. |
$45$45 - $49$49 | $50$50 - $54$54 |
Suitable for data rounded to the nearest whole number, or discrete data. |
$45$45 → $50$50 | $50$50 → $55$55 |
Not clear which endpoints are included or excluded. Assume upper endpoint is included. |
Regardless of the format used, each class interval for a given set of data should be consistent across all class intervals.
Note: In this course, class intervals for any particular set of data will be the same width. There are situations in data representation when class intervals are different widths, but this is beyond the scope of this course.
The class centre is the average of the endpoints of each interval.
For example, if the class interval is $45\le\text{time }<50$45≤time <50, or $45$45 - $50$50, the class centre is calculated as follows:
class centre | $=$= | $\frac{45+50}{2}$45+502 |
$=$= | $47.5$47.5 |
Because the class centre is an average of the endpoints, it is often used as a single value to represent the class interval. In some histograms, it may be used for the scale on the horizontal axis, with the class centre displayed directly below the middle of each vertical column.
Find the class centre for the class interval $19\le t<23$19≤t<23 where $t$t represents time.
Using the example of running times, we can add a 'class centre' column to the frequency distribution table.
Class interval | Class centre | Frequency |
---|---|---|
$45\le\text{time }<50$45≤time <50 | $47.5$47.5 | $9$9 |
$50\le\text{time }<55$50≤time <55 | $52.5$52.5 | $7$7 |
$55\le\text{time }<60$55≤time <60 | $57.5$57.5 | $20$20 |
$60\le\text{time }<65$60≤time <65 | $62.5$62.5 | $30$30 |
$65\le\text{time }<70$65≤time <70 | $67.5$67.5 | $6$6 |
The class centre is used to create an alternative to the histogram, called a frequency polygon.
A frequency polygon is a line graph that displays the frequency distribution of a set of data, and for that reason, is similar to a histogram. If we draw a frequency polygon and histogram together, the polygon begins and ends on the horizontal axis and connect the midpoints at the top of each column, as shown below.
Although a histogram looks similar to a bar chart, there are a number of important differences between them:
Histogram | Bar chart |
To better understand histograms, we will look at an example of how a histogram is created from a set of raw data.
In 2016, the World Health Organisation (WHO) collected data on the average life expectancy at birth for $183$183 countries around the world.
To appreciate what the raw data looks like, here is a reduced version of the data set.
$62.7$62.7 | $76.4$76.4 | $76.4$76.4 | $62.6$62.6 | $75.0$75.0 | $76.9$76.9 | $74.8$74.8 |
$82.9$82.9 | $81.9$81.9 | $73.1$73.1 | $75.7$75.7 | $79.1$79.1 | $72.7$72.7 | $75.6$75.6 |
$\vdots$⋮ | $\vdots$⋮ | $\vdots$⋮ | ||||
$62.5$62.5 | $72.5$72.5 | $77.2$77.2 | $81.4$81.4 | $63.9$63.9 | $78.5$78.5 | $77.1$77.1 |
$72.3$72.3 | $72.0$72.0 | $74.1$74.1 | $76.3$76.3 | $65.3$65.3 | $62.3$62.3 | $61.4$61.4 |
Each value represents the average life expectancy at birth (in years) for a single country.
Before organising the data into a frequency distribution table, we need to decide on the number of class intervals. Although there is no fixed rule, using between $5$5 and $10$10 class intervals usually produces good results for most data sets.
We know that the lowest life expectancy in the data is $52.9$52.9 years (Lesotho in southern Africa), while the highest is $84.2$84.2 (Japan). These values indicate that the scale on the horizontal axis of our histogram should be from at least $50$50 to $85$85 years. It seems appropriate to have class intervals of width $5$5 years, which means we will have $7$7 class intervals in total. We'll use the variable $t$t to represent average life expectancy in the table below:
class interval | frequency |
---|---|
$50\le t<55$50≤t<55 | $5$5 |
$55\le t<60$55≤t<60 | $10$10 |
$60\le t<65$60≤t<65 | $25$25 |
$65\le t<70$65≤t<70 | $26$26 |
$70\le t<75$70≤t<75 | $40$40 |
$75\le t<80$75≤t<80 | $49$49 |
$80\le t<85$80≤t<85 | $28$28 |
This compact table now represents all $183$183 life expectancy values.
With our frequency distribution table complete, we are ready to create a histogram:
Use the histogram to answer the following questions:
Solution
Life expectancy at birth is a measure of how long, on average, a newborn can expect to live. It is one of the most common statistics for measuring the health status of a country. The higher the life expectancy at birth, the more likely it is the country will have a high standard of living and access to quality health services and education.
As a comparison, the average life expectancy at birth for all countries in the world is $71.8$71.8 years and for Australia it is $82.9$82.9 years ($6$6th highest in the world).
In product testing, the number of faults detected in producing a certain machinery is recorded each day for several days. The frequency table shows the results.
Number of faults | Frequency |
---|---|
$0-3$0−3 | $10$10 |
$4-7$4−7 | $14$14 |
$8-11$8−11 | $20$20 |
$12-15$12−15 | $16$16 |
Construct a histogram to represent the data.
What is the lowest possible number of faults that could have been recorded on any particular day?
$\editable{}$ faults
As part of a fuel watch initiative, the price of petrol, $p$p, at a service station was recorded each day for $21$21 days. The frequency table shows the findings.
Price (in cents per litre) | Class Centre | Frequency |
---|---|---|
$120.9120.9<p≤125.9 | $123.4$123.4 | $4$4 |
$125.9125.9<p≤130.9 | $128.4$128.4 | $6$6 |
$130.9130.9<p≤135.9 | $133.4$133.4 | $5$5 |
$135.9135.9<p≤140.9 | $138.4$138.4 | $6$6 |
What was the highest price that could have been recorded?
How many days was the price above $130.9$130.9 cents?