You can change the axis scale to specify the range for the axis and whether the axis is linear or transformed. These options apply only to scale axes.
How to Change the Axis Scale
Note: You can also use the rescaling tool to change the scale quickly. However, this tool changes the scale of all axes in the chart. See the topic Changing Axis Scales for more information.
- Select a scale axis. Note that there are separate steps to change the axis scale on a matrix scatterplot. See the topic Axis Scale for Matrix Scatterplots for more information.
- If the Properties window is not already displayed, from the menus
- Use the Scale tab to specify the axis scale options.
- Click Apply.
Using the Scale Tab
If Auto is selected for the values, the Chart Editor automatically sets the value to include all of the data and to create appropriate tick labels (for example, 10 instead of 10.0123). If you don't get the exact results that you expect from the range settings, deselect Auto for each setting and enter custom values.
Minimum/Maximum. Change the range for the axis. The minimum and maximum values in the data for the graph are listed so that you can set a range that includes all of the data. Other graph elements (for example, annotations) can be hidden when you change the range. If elements you expect to see aren't displayed, change the range back to automatic. If the scale is transformed, the range values are specified in the same units as the data values.
Major Increment. Specify the size of the increments between major ticks/labels. Beginning with the minimum value, a major tick occurs at every increment specified by this number. In general, a division size that divides evenly into your range works best. For example, if your axis minimum is 0 and the maximum is 400, division sizes of 100, 50, or 25 would work well.
Origin. Specify the origin. The origin has a different effect depending on the graph type.
- For charts with bars and lines, the origin specifies the line from which bars and areas extend. The graphic elements originate at the origin and extend toward their value. For example, if your bar chart includes values of 367 and 48 and the origin is 100, one bar extends up from 100 to 367 (in default coordinates), while the other bar extends down to 48. The specified value for the origin must fall within the minimum and maximum.
- For a scatterplot, specifying an origin is useful only with a transformed axis. The effect of changing the origin is most obvious when the data are far from the origin, because the transformation takes the origin into account. For example, assume that the axis is a log scale. The equation for the transformation is y'=log(y-origin). Therefore, if the data points are 1001, 1002, 1010 and the origin is 0, the transformation will look similar to an untransformed scale. However, if you change the origin to 1000, the transformation will look like a log scale as expected.
Display Line at Origin. Displays the axis at the value specified in the Origin text box.
Type. Displays the data on a linear or transformed scale. Scale transformations help you to understand the data or to make assumptions necessary for statistical inference. On scatterplots, you might use a transformed scale if the relationship between the independent and dependent variables is nonlinear. You can often transform these relationships into a straight-line relationship. Scale transformations can also be used to make a skewed histogram more symmetric and more like a normal distribution. Note that you are transforming only the scale on which the data are displayed; you are not transforming the actual data.
- Linear. Displays a linear, untransformed scale.
- Log. Displays a log-transformed scale. The formula for this transformation
log(x). Optionally, you can enter a base for the log, which must be greater than 1. If Safe is selected, the transformation formula used for taking the log of an axis value x is not
log(x). The Chart Editor uses a different formula (safe log), so that it can handle 0 and negative values.
Formula for Safe Log Transformation
The safe log formula is:
sign(x) * log(1 + abs(x))
So if you assume that the axis value is –99, the result of the transformation is:
sign(-99) * log(1 + abs(-99)) = -1 * log(1 + 99) = -1 * 2 = -2
- Power. Displays
an exponent-transformed scale. The formula for the transformation
power(x, exponent). Optionally, you can enter an exponent value. The default exponent is 0.5, which would take the square root of the data.
- Logit. Displays a logit-transformed scale. The formula for the transformation
log(1/(1-x)). Data values for this scale must fall in the open interval (0, 1). That is, for any data value x, 0 < x < 1.
- Probit. Displays a probit-transformed scale. The formula for the transformation is the inverse cumulative distribution function (CDF) of the normal distribution. Data values for this scale must fall in the closed interval [0, 1]. That is, for any data value x, 0 ≤ x ≤ 1.
- Inverse Sine. Displays an inverse sine-transformed
scale (also called an arcsine scale). The formula for the transformation
arcsin(x). Data values for this scale must fall in the closed interval [0, 1]. That is, for any data value x, 0 ≤ x ≤ 1.
- Hyperbolic Arctangent. Displays a hyperbolic
arctangent-transformed scale (also called an inverse hyperbolic tangent
scale or a Fisher's z scale).
The formula for the transformation is
arctanh(x). Data values for this scale must fall in the open interval (-1, 1). That is, for any data value x, -1 < x < 1.
- Complementary log-log. Displays a complementary log-log-transformed scale (also called
a Weibull scale). The formula for the transformation is
log(log(1/(1-x)))Data values for this scale must fall in the open interval (0, 1). That is, for any data value x, 0 < x < 1.
Margins. Creates a margin around the data. You specify a percentage (0 to 50) of the inner frame to use for the margin. The margin appears perpendicular to the selected axis. For example, if you set the upper margin to 5% for the vertical axis, a margin whose thickness is 5% of the data area runs along the top of the data frame.