· Equivalence is not "no difference": you first set an "acceptable equivalence interval" (±Δ or ±%), then use TOST (two one-sided t-tests) to show the mean difference falls inside it; both TOST p-values must be <α to declare "equivalent".

· 2 independent samples use Welch-Satterthwaite df; paired uses the t-test on differences; one-sample compares against the reference. The ratio mode tests on the log scale (Δ%, e.g. entering 10 means an equivalence interval of ±10%, i.e. 90%~110%; bioequivalence commonly uses ±20% or 80-125%).