The following steps generate a simple line chart displaying a single timeline:

Figure 10.1. A line chart displaying the mean current salary for the year of birth of the employee.

You can have more than one line appear on a chart by adding more than one variable name to an axis. But the variables must contain a similar range of values before they can be represented by the same axis. For example, if one variable ranges from 0 to 1,000 pounds and another variable ranges from 1 to 2 pounds, the values of the second variable will show up as a straight line, regardless of how much it actually fluctuates.

The following steps generate a multiline graph:

Figure 10.3. A line graph charting two variables.

The variables you choose as members of the Y-axis must have a similar range of values to make sense. For example, if you were to choose age and annual income as two variables to be charted together, the result wouldn't be all that interesting; the salary values would be in the thousands and the ages, regardless of their variation, would all appear in a single line.

The following steps show you how to construct a simple scatterplot:

Figure 10.4. A simple scatterplot showing the effect of starting salary on current salary.

Each dot on the scatterplot in Figure 10-4 represents both the starting salary and the current salary of one employee. The most obvious fact you can derive from this is that the current salary depends largely on the starting salary. In the pattern of the dots, it's easy to see a normal line from the lower left to the upper right. Any dot on that imaginary line represents the salary of an employee who received a normal raise. The dots above the line are the employees who got above-average raises, and those below the line are those with below-average raises. This plot has the shortcoming that the length of service is not considered.

You can display the values of more than one variable along the same axis. The following example constructs a scatterplot showing the beginning salary and the current salary according to the number of months of experience the person had before taking the job:

Figure 10.5. A scatterplot showing salary information organized by previous experience.

Three-dimensional scatterplots can be dramatic in appearance, but clarity is not their strongest point. Because the scatterplot is drawn on a two-dimensional surface, you might find it difficult to envision where each point is supposed to appear in space. On the other hand, if your data distributes appropriately on the display, the resulting chart may demonstrate the concept you're trying to get across.

The following example uses the same data as in the preceding example but displays it in a different way, as a three-dimensional plot:

Figure 10.6. A three-dimensional scatterplot printed on a two-dimensional surface.

Three-dimensional scatterplots can have more than a single value along an axis. However, as in the previous example, clarity is not the strongest point of the result. Depending on the data, you might find it difficult to envision where each point is supposed to appear in space — or the chart may demonstrate the concept you're trying to get across. All you can do is generate the graph and then evaluate whether or not you want to use the result.

The following example uses the same data basic information that was displayed in the preceding example:

Figure 10.7. A grouped three-dimensional scatterplot printed on a two-dimensional surface.

A Summary Point plot has the same layout as a bar chart, except the bars are not drawn. Instead, a point is placed at the position of the top of the bar. This means that it is possible to place more than a single value at the position that would otherwise be filled by a single bar. The points for different data items vary in appearance, and are identified by a legend. You must be careful, however, to choose variables that make sense when combined and displayed this way.

The following example shows the relationship between the means of cubic-inch engine displacement and the mean of automobile horsepower over several years:

Figure 10.8. A Summary Point plot showing the difference of the trends of two variables.

No plot is simpler to produce than the dot plot. It has only one dimension. Although SPSS groups it among the scatterplots, there's nothing scattered about it. It actually presents data more like a bar chart — and it reminds me of that old joke about stacking BBs.

It's easy to create a dot plot. You select the dot plot as the type of graph you want and then select one variable. SPSS does the rest. The following steps guide you through the process of creating a simple dot plot:

Figure 10.9. A dot plot showing the relative numbers of persons born in each year.

A scatterplot matrix is a group of scatterplots combined into a single graphic image. You choose a number of scale variables and include them as a member of your matrix, and SPSS creates a scatterplot for each possible pair of variables. You can make the matrix as large as you like — its size is controlled by the number of variables you include.

The following steps walk you through the creation of a matrix:

The matrix of scatterplots in Figure 10-10 has each variable plotted against each of the others. Notice that the scatterplots along the diagonal from the upper left to the lower right are blank — that's because it's useless to plot a variable against itself. Also, notice the symmetry: Each plot in the lower-left half has a rotated and mirrored image in the upper-right half.

Figure 10.10. A four-by-four matrix of scatterplots.

A drop-line chart presents a special kind of summary with points and vertical lines. The points are grouped horizontally; at each categorical value, a line is drawn vertically through them. This arrangement can be visually helpful for comparing the values that appear within each category.

The following steps take you through the basic actions necessary for producing a drop-line graph:

Figure 10.11. A drop-line plot is based on a pair of categorical values and a scale variable.

A fundamental bar graph is simple enough that the decisions you need to make when preparing one are almost intuitive. The following steps can be used to generate a simple bar graph:

Figure 10.12. A simple bar graph.

A clustered bar chart can show the relationships among a cluster of items by displaying more than one value and presenting a summary of categorical values. Clustering combines several bar charts into one. The following steps take you through the process of constructing a clustered bar chart:

Figure 10.13. A bar graph with values displayed as clusters.

A stacked bar chart is similar to the clustered bar chart in that it displays multiple values of a variable for each value of a categorical variable. But it does it by stacking them instead of placing them side by side. The following chart displays the same data as the preceding example, but emphasizes different aspects of the data.

Following these steps will create a stacked bar chart:

Figure 10.14. A bar graph with values displayed in stacks.

A simple three-dimensional bar chart is the same as a two-dimensional bar chart, except a third variable is added to specify the values along the new dimension. As with most three-dimensional displays, it has the advantage of displaying three relative values at once, and it has the disadvantage of making it difficult to determine which is the greater of two values if the two values are close. You'll just have to try it with your data and decide whether it demonstrates what you've got or just contributes to the general confusion.

The following steps construct a three-dimensional bar chart:

Figure 10.15. A three-dimensional bar graph.

A clustered three-dimensional bar chart can show the relationships among a cluster of items by displaying more than one value and presenting a summary of categorical values. Clustering combines several bar charts into one. A great deal of information is displayed in a single chart, and it can be confusing in appearance. You will need to test it with your data and determine whether it demonstrates what you are trying to show.

The following steps take you through the process of constructing a clustered bar chart:

Figure 10.16. A bar graph with values displayed as clusters.

A stacked three-dimensional bar chart is very similar to the clustered bar chart. The two can be used to show the same data, but in slightly different ways. A stacked chart can show the relationships among data items by stacking more than one value and presenting a vertical summary of categorical values. Stacking combines several bar charts into one. Because that single chart displays a great deal of information, however, it can be confusing in appearance. The layout may even hide some of the data inadvertently. You'll have to test it with your data to see whether it shows what you want to see.

The following steps take you through the process of constructing a stacked three-dimensional bar chart:

Figure 10.17. A bar graph with values displayed in stacks.

Some errors come from flat-out mistakes — but those aren't the errors I talk about here. Statistical sampling can help you arrive at a conclusion, but that conclusion has a margin of error. This margin can be calculated and quantified according to the size of the sample and the distribution of the data. For example, suppose you want to know how typical the result is when you calculate the mean of all values for a particular variable — for any one case, the mean could be as big as the largest value or as small as the smallest. The maximum and minimum are the extremes of the possible error. You can choose values and mark the points that contain, say, 90 percent of all values. Marking these points on graphs creates error bars.

You can add error bars to the display of most types of graphs. For example, you could add error bars to the simple bar graph presented earlier in this chapter (refer to Figure 10-12) by making selections in the Element Properties dialog box. If you've worked through any of the examples, you'll know Element Properties as that pesky window that pops up every time you construct a chart.

For an example of adding error bars to a bar chart, follow the same procedure described previously in the "Simple bar graphs" section — but just before the final step (clicking the OK button to produce the chart), do the following:

Figure 10.18. A bar chart with error bars that show a range containing 95 percent of all values.

Another way to display the same data is to display the range of errors without displaying the full range of all values. To do so, follow these steps:

Figure 10.19. An error bar graph, showing the mean values as dots and the upper and lower bounds of the error.

This example displays the result of one way of making error calculations: The magnitude of the error is based on 95 percent of all values being within the upper and lower error bounds. If you prefer, you can base the error on the bell curve and mark the upper and lower errors at some multiple of the standard error or standard deviation.

A clustered error bar chart can show the relationships among a cluster of error ranges by presenting a summary of categorical values. Clustering combines several error bar charts into one. The following steps take you through the process of constructing a clustered error bar chart:

Figure 10.20. An error bar graph, showing error ranges displayed as clusters.