Calculating the number of répétitions
Ellistat provides a little tool to calculate the number of repetitions necessary depending on the number of factors and interactions in the selected table and the smallest effect that we want to highlight in regards to the expected residual dispersion from the tests.
Example: you want to create an design of experiment with an L8 table to study 5 factors and 2 interactions.
The study result is the temperature obtained. The target is 120°c ± 10°c.
Any factor for which the effect is less than 1° does not interest us. So we enter 1 for the smallest effect that we wish to highlight.
We know that normal production gives us a temperature that varies ± 4° (estimation made by experts). This estimation covers 95% of the real dispersion (± 2 sigma) so we estimate the residual standard deviation as 2.
If we know the short term standard deviation of the process from another study, we enter this standard deviation rather than making an estimate.