Differences between means, proportions, and rates can theoretically range from negative infinity to positive infinity, with their expected values, assuming no association, being 0.

If one of the means (or proportions) being compared may theoretically be either infinitely less than or infinitely greater than the other one, and if the two means (or proportions) are equal, then the difference between them will be 0.

The distribution of a difference measure is about normal if the sample size is reasonably large. Statistical testing, therefore, can usually rely on the known values of the standard normal curve.