By Kevin Beck, Updated Aug 30, 2022
Imagine you want to know how your 12‑week‑old purebred puppy’s weight stacks up against other dogs of the same age, sex, and breed worldwide. If you have access to a comprehensive database, you can compare your puppy’s weight to the population average and see how it ranks. But what if you only have a handful of data points and still want to gauge how a particular value relates to the broader population?
In such cases, two statistical tools come into play: the z‑score and the t‑score. Both help you understand how a specific observation compares to a “typical” value, but they are used under different circumstances.
The mean (average) of a data set is the sum of all values divided by the number of observations, n. For a population, the mean is denoted by μ, and the standard deviation by σ. In a standard normal distribution, about 68% of observations lie within ±1 σ of the mean, and about 95% lie within ±2 σ.
The magnitude of the standard deviation relative to the mean indicates the spread of the data: a larger σ produces a wider bell curve, while a smaller σ results in a narrower one.
A z‑score measures how many standard deviations a single observation, x, is from the population mean: Z = (x – μ) / σ. A z‑score of 0 means the observation equals the mean; +1.00 and –1.00 indicate one standard deviation above or below the mean, respectively.
A t‑score is similar but uses the sample mean (𝑥̄) and the sample standard deviation (s), and incorporates the sample size: t = (𝑥̄ – μ) / (s / √n). The denominator represents the standard error of the mean.
If your sample contains fewer than 30 observations, a t‑score is preferred over a z‑score. As the sample size grows, the t‑distribution converges toward the normal distribution, making the difference negligible for large n. The choice of confidence interval—typically 90% or 95% for two‑tailed tests—determines the critical value you compare your t‑score against.
Suppose a class of 25 university students averages 64% on a surprise Harry Potter trivia test. The population mean is 60% and the sample standard deviation is 15%. To compute the t‑score:
t = (64 – 60) / (15 / √25) = 4 / (15 / 5) = 4 / 3 ≈ 1.33
The degrees of freedom are df = n – 1 = 24. Looking up a 90% confidence level in a t‑distribution table (or using an online calculator), the critical value for 24 df is about 1.711. Since 1.33 < 1.711, the class average is not significantly higher than the population mean at the 90% confidence level.
Adjusting the confidence interval (e.g., to 80% or 70%) would change the critical value and could alter the conclusion.
For more detailed tables and calculators, consult reputable sources such as the Wikipedia entry on the t‑distribution or statistical software like R or Python’s SciPy library.
