Lenth's pseudo standard error

significance of effects without replicates

PSE = 1.5 · median(|c| : |c| < 2.5 · s₀), s₀ = 1.5 · median|c|

Russell Lenth's method (1989) estimates the noise from the effects themselves, assuming that most effects in a screening experiment are small. Effects that stand out from the pseudo standard error by more than a t-quantile with m/3 degrees of freedom are judged significant.

It is used for unreplicated factorial designs and works best with 8 or more effects, so from 3 factors up.

Calculate it

DOE: 2-Level Factorial Design: 2^k factorial: run plan, effects, Pareto, interaction plots and best settings.

Related terms

• Pure error: noise estimated from replicated runs
• Pareto of standardised effects: effects ranked by size against a significance line