Degrees of Freedom and the Conservative Rule

When you compare two means with a t procedure, the exact degrees of freedom formula blends both sample sizes and both sample standard deviations into one ugly decimal. Calculators compute it instantly, but by hand it is painful. The conservative rule dodges this: take the smaller of the two sample sizes, call it n1, and use df = n1 - 1. That's it. No variances, no decimals.

Why does this work as a safe shortcut rather than a cheat? Smaller df makes the t-distribution wider and more spread out, which pushes the critical t-value up. A bigger critical value means a wider confidence interval and a larger margin of error. So using the smaller sample's df minus 1 never underestimates how uncertain your estimate is — it errs toward caution. The real formula can never give a df larger than the smaller sample size minus 1, and it can never be smaller than the smaller of (n1-1) and (n2-1) either, so the conservative value is a guaranteed floor. It is not n1 + n2 - 2, because that pooled-df formula belongs to a different procedure, one that assumes the two population variances are equal — an assumption the two-sample t for means does not make.

As both samples grow, the exact df formula and the conservative df both climb toward the same neighborhood, so the intervals they produce get closer together. This matters going forward: the same tension between an exact method and a simplified, safer stand-in shows up whenever software and hand calculations disagree slightly — always trust the more exact value when you have it, and treat the conservative rule as a backup or a sanity check.