topic Re: The same run in doe, analysis of several results in Discussions

The same run in doe, analysis of several results

YanivD — Fri, 09 Jun 2023 00:54:21 GMT

Hi,
I have doe with 12 runs.
In each run, for the same response i have 5 measured samples.

How to analyze the data?
To enter all the result, i.e. 5 rows for each run.. Or any other efficient way?

Thanks

Re: The same run in doe, analysis of several results

louv — Sun, 04 Sep 2022 11:49:55 GMT

If you have 12 runs, you should have 12 rows and the 5 measurements should be added next to the rows. If you add them as rows you will be treating them as if they were independent experiments whereas from your question it appears that you have only 12 experiments.

Re: The same run in doe, analysis of several results

statman — Sun, 04 Sep 2022 13:50:26 GMT

Are the 5 measurements the same response variable? Assuming you are measuring the the same Y's, you have 1 experimental unit for each treatment and it consists of 5 repeated measures. Follow repeated measures analysis.

Re: The same run in doe, analysis of several results

YanivD — Sun, 04 Sep 2022 14:52:04 GMT

Thanks, for each run i am measuring the diameter of the part. I have 5 the same parts with little deviation between the diameter.. Again all came from the same run factors.
What should I do?

Re: The same run in doe, analysis of several results

louv — Sun, 04 Sep 2022 16:33:23 GMT

You can add a column for the mean diameter and standard deviation for the measurement as well. I assume you have a diameter target.

Re: The same run in doe, analysis of several results

statman — Mon, 05 Sep 2022 19:06:17 GMT

My thoughts:

What you do depends on how the actual data was gathered. It is possible you have multiple components confounded in the measurements:

1. Part-to-part,

2. Within part, and

3. Measurement

I would first look graphically at the within treatment and possibly assess the consistency within treatment. Then, as suggested calculate appropriate enumerative statistics to summarize the within treatment data (mean and variance) and then model both response variables.