Calculate Disparity Statistics

To start, choose Disparity from the Methods menu.  The disparity statistic can be applied to datasets in both polygon and point geogaphies.  For SpaceStat to calculate the disparity statistic, it needs a disease rate or count, and a population dataset for the reference and target populations, the rate multiplier (if inherent in your disease rate data), the time(s), how to correct for multiple tests, and your input on how to name the results folder.

The reference population is the one to which you are comparing the target population. If you're interested in whether, say, cancer rates in black females are different from rates in white females, based on the CDC's suggested standards for disparity analyses, you would look to see which population had the most favorable rate (in this case, white females), and use this population as the reference.

Click within the sections in the images below for more information.

Input

When the task manager opens for calculating the Disparity statistic, it will start on the "Input" section.  Here you will enter your reference and target datasets, choose which disparity statistics to generate (we recommend doing both, so you can compare--strong differences are most likely with datasets for events like rare cancers, which can have unstable rates due to variation in the population sizes across your geography), and choosing the start and end times and rate multiplier.  As a default, the rate multiplier is set at the value used by the National Cancer Institute for expressing cancer mortality rates (X deaths per 100,000 people).  Make sure to change this value if it does not apply to your data.

Advanced

The advanced section for the Disparity statistics asks you to choose between two-tailed (do the two rates differ, either in absolute or relative magnitude?) and one-tailed tests (does the absolute or relative difference in rates exceed a particular threshold?).  In addition, you can change the alpha level, which is one way of addressing the problem of multiple testing, or choose the FDR or extended Simes methods to correct for multiple tests. If you select a one-tailed test, you then also need to indicate the relevant thresholds for each statistic.  Note that the difference threshold should be in the same units as your dataset (i.e., do not account for the rate multiplier, if there is one), and the ratio threshold should be a ratio, not a percentage.  The 0.10 value entered below means that the statistic would test for a 10% exceedance in the target rate relative to the reference.  Values for either can be zero, so you can simply test whether the target rate exceeds the reference rate (one-sided test, rather than testing if the two differ).  The threshold (one-sided) tests implemented in SpaceStat are designed for situations where the target rate exceeds that in the reference population (a standard approach suggested by the CDC), so by threshold values must always be positive.  The form of data, and choice of Reference and target populations should thus be chosen to make the difference positive.  

Output

The output section of the task manager for the Disparity statistic asks for name for the output folder.  The default name is "Disparity"; here we have made this name a bit more descriptive.

Run method

The run method page presents a summary of the previous three pages so that you can review the choices you have selected.  If everything here looks correct, select the "Run" button at the bottom of the page.   The map output from this run is shown in the image on the "Disparity Results" page, and map output from the two-tailed test for the same datasets is shown on the "About the Disparity Statistics" page.