topic Re: Easy: Determine Main Effects in Discussions

Easy: Determine Main Effects

Eric_Johanson — Sat, 10 Jun 2023 23:29:22 GMT

After a full factorial design was executed I have the following tabulated results:

Force	Pressure	Hours to failure
15 N (Low)	10 Pa (Low)	2800 hours
15 N (Low)	150 Pa (High)	1300 hours
100 N (High)	10 Pa (Low)	800 hours
100 N (High)	150 Pa (High)	300 hours

How would I determine the main effects of Force and Pressure? I've added this data to a table, my next move would be to Analyze > Fit Model and figure out to make a Least Squares Mean Plot. Having a hard time setting that up in the Fit Model.

Re: Easy: Determine Main Effects

txnelson — Wed, 28 Apr 2021 18:11:03 GMT

Did you only measure one Hours to failure for each combination of Force and Pressure?

Re: Easy: Determine Main Effects

Eric_Johanson — Wed, 28 Apr 2021 18:15:49 GMT

Yep, you got it!

Re: Easy: Determine Main Effects

txnelson — Wed, 28 Apr 2021 18:24:31 GMT

Enter your data into a data table

Go to

Analyze=>Fit Model

enter in the model you want to analyze

And click OK

Re: Easy: Determine Main Effects

statman — Wed, 28 Apr 2021 19:12:56 GMT

Eric, welcome to the community. Jim provided the steps to do exactly what you want. I would simply add...why not also get the interaction effect while you're at it. 3 total DF's P+F+PF

Re: Easy: Determine Main Effects

txnelson — Wed, 28 Apr 2021 20:14:03 GMT

@statman

Not enough df

Re: Easy: Determine Main Effects

statman — Wed, 28 Apr 2021 20:33:14 GMT

Ummm, yes there are. I always start with a saturated model, use Normal, Bayes and Pareto plots as well as look at parameter effects and SS. Use engineering knowledge and predictions about potential effects to then reduce the model. I really don't place much weight on p-values as YOU control these via how you collect your data. So, for example, if you leave the interaction out of the model, guess what...the interaction is the estimate of mean square error and is therefore the basis of the statistical test (F-test). For unreplcated designs, I much prefer the unbiased Daniels plots, but that's just me.

Re: Easy: Determine Main Effects

txnelson — Wed, 28 Apr 2021 20:52:55 GMT

I stand corrected.....