# Power and sample size for factorial design

F

Hey guys.
I have a series of data for a "2 level full factorial design" for 4 factors.

After analyzing the data, I want to run the POWER AND SAMPLE SIZE for that which requires standard deviation as an input data.
minitab help says: For 2-Level Factorial Design use the square root of the mean square error (MS Error).

My problem is to calculate square root of the mean square root in minitab. which tab/part of the program gives me the answer

If any further data is required, please inform me.

#### Miner

##### Forum Moderator
Look for the ANOVA table and locate the row called Residual Error. Follow this row across to the column entitled Adj MS. This number is the MS Error.

You can also use the standard deviation from a capability study or from SPC charts if those are available.

#### Statistical Steven

##### Statistician
Super Moderator
Unless you ran repeats your power will be very low. Even with repeats, DOE is not intended to look for small effects, but rather find larger effects so the power is low. FYI

Hey guys.
I have a series of data for a "2 level full factorial design" for 4 factors.

After analyzing the data, I want to run the POWER AND SAMPLE SIZE for that which requires standard deviation as an input data.
minitab help says: For 2-Level Factorial Design use the square root of the mean square error (MS Error).

My problem is to calculate square root of the mean square root in minitab. which tab/part of the program gives me the answer

If any further data is required, please inform me.

#### Miner

##### Forum Moderator
Unless you ran repeats your power will be very low. Even with repeats, DOE is not intended to look for small effects, but rather find larger effects so the power is low. FYI

I think you mean replicates. In DOE there is a distinction between repeats and replicates.
Repeats are multiple measurements taken during the same experimental run.
Replicates are multiple experimental runs under the same factor levels.
Replicates increase statistical power and include setup to setup variation in the residual error.

#### Statistical Steven

##### Statistician
Super Moderator
I think you mean replicates. In DOE there is a distinction between repeats and replicates.
Repeats are multiple measurements taken during the same experimental run.
Replicates are multiple experimental runs under the same factor levels.
Replicates increase statistical power and include setup to setup variation in the residual error.

You are correct that replicates increase statistical power (especially in Minitabs calculation). I am speaking of also getting a better estimate of the MSE by taking multiple readings (repeats).