Page:Sm all cc.pdf/55



Plotting Hints
Crossplots can hide or reveal patterns. Plotting technique affects the efficiency of visual pattern recognition. Scientists are accustomed to a suite of plotting conventions, and they may be distracted if asked to look at plots that depart substantially from these conventions. I thank Open University [1970] for reminding me of some of the following plotting hints, which I normally take for granted. Figure 12 illustrates the effect of a few of these factors.
 * Plot the dependent variable (the one whose behavior you hope to predict from the other variable) on the vertical axis, and plot the independent variable on the horizontal axis.
 * Choose an aspect ratio for the plot that maximizes information (e.g., if we are examining the changes in Y values throughout a long time series, then the horizontal X axis can be much longer than the vertical Y axis).
 * Plot variables with values increasing to the right and upward.
 * Choose simple scale divisions, usually with annotated major divisions and with tics for simple subdivisions (e.g., range of 20-40 with annotation interval of 5 and tic spacing of 1).
 * Choose a total plot range for each variable that is as small as possible, subject to these two restrictions: simple scale divisions and inclusion of all data points.
 * Make an exception to the previous hint by including major meaningful scale divisions such as zero or 100%, only if this inclusion requires a relatively small expansion of the plot range.
 * Plot data points as solid or open circles or crosses.
 * If more than one dataset is included on the same plot, use readily distinguishable symbols.
 * Label each axis with the variable name and its units.
 * If data are a time series, connect the points with line segments. If they are independent, fit a line or curve