Response Info
Responses are the quality characteristics being measured at each run in the experiment. These must be numeric, and the values must be sensitive enough to reflect changes in the process or product as the run conditions are changing.
Responses: The number of responses being measured.
Edit Model: The default power model is for all possible main effects, click to change that setting.
Edit response types [on]: Unchecking this box will revert all the response distributions to continuous and turn off the proportion only information fields.
Options: Click to change which formulas are used to compute the signal to noise ratio for proportion responses.
Response Information
Name: A descriptive name for the response.
Units: The units of measure for the response.
Response Type: Choose Continuous or Proportion.
Continuous is used for a standard measured response.
Proportion is used when the response is similar to fraction defective.
Continuous Response
Diff. to Detect (Difference to Detect) or Delta “Signal” is the minimum change in the response considered large enough to be a practical change. It is the minimum amount of change you want the design to detect as significant. A larger change is easier to detect resulting in higher power.
Est. Std. Dev. (Estimated Standard Deviation) or Sigma “Noise” is an estimate for the run to run standard deviation of the response. The estimate can come from a pilot study, a similar process, scientific knowledge and as a last resort – guess.
The Delta/Sigma “Signal/Noise” ratio is used to determine the probability of detecting an effect.
The default “Signal” of 2 and “Noise” of 1 are a common setting for checking power. Leaving these as is will estimate the power for a 2 standard deviation effect. Change these values to something more realistic to get a better estimate of power for the experiment.
Proportion Response
Diff. to Detect (Difference to Detect) or Delta “Signal” is the minimum change in the proportion considered large enough to be a practical change. It is the minimum amount of change you want the design to detect as significant. A larger change is easier to detect resulting in higher power.
Samples per Run: To get a representative sample to estimate the true proportion at each run’s setting, several samples are necessary. A larger, per-run sample will improve the power.
(p-bar) is the best estimate for the average of the proportions that will be produced during the experiment. A good estimate is the current operating output proportion.
Run Variation (% of p-bar) is the percentage the proportion is expected to vary between replicate runs.
The Delta/Sigma “Signal/Noise” ratio is used to determine the probability of detecting an effect.
Press the Next button to calculate power and/or move on through the build.