About the Index of Dispersion
The Index of Dispersion Calculator computes the variance-to-mean ratio of a set of counts to check whether they're distributed more randomly, more clustered, or more evenly than a Poisson process would predict. It's a quick screening tool used in fields like ecology and epidemiology, where count data such as organisms per plot, cases per region, or defects per batch is common.
How It Works
You enter a list of counts, such as the number of events observed in each of several equal-sized intervals or locations. The calculator finds the sample mean and sample variance of those counts, divides variance by mean to get the index of dispersion, and labels the result as overdispersed, underdispersed, or consistent with random variation.
Formula & Methodology
For a true Poisson process, variance equals the mean exactly, so the ratio should sit near 1. The calculator flags a ratio above 1.2 as overdispersed, meaning the counts cluster more than chance alone would produce, and a ratio below 0.8 as underdispersed, meaning the counts are more evenly spread than random chance would allow. Ratios between those thresholds are treated as consistent with random, Poisson-like variation.
Examples
Clustered Event Counts
Ten intervals show counts of 3, 5, 2, 8, 4, 1, 6, 3, 9, 2. The mean is 4.3 and the sample variance is about 7.12, giving a variance-to-mean ratio of roughly 1.66, which the calculator flags as overdispersed, more clustered than a random process would produce.
Tightly Regulated Output Counts
A production line reports hourly output counts of 4, 5, 3, 6, 4, 5, 4, 3, 5, 6. The mean is 4.5 and the sample variance is about 1.17, giving a ratio of roughly 0.26, which the calculator flags as underdispersed, consistent with a tightly controlled process.
Advantages
- Gives a single, easy-to-compute number to screen whether count data behaves like a random Poisson process before applying more complex models.
- Labels the result in plain language, overdispersed, underdispersed, or random-like, rather than leaving interpretation entirely to the user.
- Works with any list of raw counts, without needing to know the underlying process that generated them.
Common Mistakes
- Treating a ratio just above or below the 1.2 / 0.8 thresholds as statistically conclusive, when it's really a rough guideline rather than a formal significance test.
- Applying the index of dispersion to data that isn't count data, such as continuous measurements or percentages, where the Poisson-based interpretation doesn't apply.
- Drawing conclusions from a very small number of counts, where sample variance is too unstable to reliably indicate over- or underdispersion.
Edge Cases to Watch For
- At least 2 counts are required, and the mean of the counts cannot be zero, since dividing by a zero mean is undefined.
- The 1.2 and 0.8 thresholds used to label overdispersion and underdispersion are practical guidelines built into this calculator, not a formal statistical hypothesis test with a p-value.
- Small sample sizes produce a noisy variance estimate, so a ratio that looks overdispersed or underdispersed from just a handful of counts may not hold up with a larger sample.
Common Use Cases
- Ecologists checking whether organisms are randomly distributed across sample plots or instead clustered or evenly spaced.
- Epidemiologists screening disease case counts across regions or time periods for unusual clustering before deeper analysis.
- Quality control analysts checking whether defect or event counts across production batches are more variable than expected by chance.