4 Making sense of data
Histograms are a special form of bar chart in which the bars usually touch each other because histograms always show data collected into ‘groups’ along a continuous scale. They tend to be used when it's hard to see patterns in data, for example when there are only a few variables, or the actual amounts are spread over a wide range. For example, suppose you manufactured biscuits; it is important to manufacture closely to a given size, as there are regulations governing the sales of biscuits that are not the advertised size. By grouping your data along a continuous scale measuring size, you can see quickly whether the size is near to the ideal, or whether your machinery needs to be adjusted.
We would now like you to look at an example based on engineering pay settlements and answer the questions that follow. You will notice that the bars don't actually touch, which they should do.
- What is shown on the vertical axis?
- What was the most common pay settlement?
- How many settlements were for a pay freeze?
- Where was the data in the histograms derived from?
- What specific influences has the data source had on the graphs, diagrams or articles and its conclusions in particular?
- How does it relate to the article?
- What conclusions can you draw about the use of the histograms and related text?
Technically, the group sizes in a histogram should all be the same. However, the authors have chosen to group some of the pay rises together, probably to keep the graph small.
- The percentage of engineering firms in the sample.
- The most common settlement was between 3 per cent; and 4 per cent; 57 per cent of the sample gave rises in that range.
- The graph shows that 10 per cent of settlements were for a pay freeze. The article says that 10 firms gave this.
- The Engineering Employers Federation, an employers organisation representing some of the country's engineering firms, is the source of the data. (The Federation has a series of regional offices, with a Head Office in London, which is presumably where the data came from. Member firms pay a levy to join the EEF. It isn't clear from the histogram or the accompanying article whether the EEF have surveyed only their own members or engineering firms in general.)
- As with the articles on the pie charts, the magazine seems to have been influenced by the people who have provided the data, to the extent of mentioning their name frequently. In this case, the EEF have had a telephone number printed as well.
- It relates well to the article, which talks about the spread of pay. The authors have also added two other pieces of information though. First, they talk about the pressure downwards on pay settlements, which is not in the diagram. In addition, they talk about the number of deals involved.
- If you add up all the percentages, you will see that they come to 105 per cent. This isn't unusual, as charts often show rounded figures, for ease of understanding, although it does seem a lot. However, look at the article. You will see that they say that 8 deals were for over 4 per cent (which includes 7 per cent of deals in the 4 per cent to 5 per cent range, and 1 per cent of deals in the over 6 per cent range) and they say that 10 deals were for a pay freeze. There are 10 pre cent of deals in the pay freeze range. Therefore, it seems odd that the other numbers don't relate quite as neatly to the formula: one deal =1 per cent.
In addition, again as a small point, we found the format of the graph to be distracting. It's very interesting and it's obviously a relevant drawing in the background, but it perhaps makes it hard to look at the actual figures.