evaluation of DOE experiment

[#1]

Hello,

I established a randomised definitive Screening design using the corresponding menue point under DOE and I created the data table and saved it some month ago. Now the experiments will be finished soon.

How can I add the result data and analyse the results?

I don't see how to use and load the data table of the design I created before. There is no project with the entered data that I could simply reopen. It is not possible to reproduce the same order of the runs in order to create exaclty the same design once more. Import of the design data from an Excel sheet is not possible.

How must I proceed to use the results with the corresponding data table of the design?

Sincerely

Thommy7571

[#2]

Did you save the resulting JMP data table when you created the design in the first place? And importing an Excel worksheet into JMP can be as simple as a copy/paste from Excel into an empty JMP data table.

[#3]

Hallo,

<<Did you save the resulting JMP data table when you created the design in the first place?

yes, I saved the design data as I mentioned as a data table as I already mentioned.

<<and importing an Excel worksheet into JMP can be as simple as a copy/paste from Excel into an empty JMP data table.

Someone told me that due to meta data this is not possible without problems

How to proceed now?

Sincerely

Thommy7571

[#4]

Hi @DrThWillms,

In order to export/import JMP files as Excel without losing metadata (like column properties or other informations), I would recommend using the Table Attributes Add-In. 

There is a post explaining how to use it : JMP to Excel and back again -- all without losing any table attributes 

This might be particularly helpful if you share the design through Excel file to collect the response measurements, and want to proceed with the analysis without losing all the scripts and column properties already set in JMP.

Hope this can help you,

[#5]

HI @DrThWillms ,

  There's a couple ways you can go about this.

1. Do as @P_Bartell mentions and use the DOE data table that you originally created (and hopefully saved -- from your post, you do mention that you saved it) to then run the experiments. You can then manually enter the data in JMP (or copy/paste from Excel), or read in the Excel file as was mentioned and copy/paste.

2. You can also always set a random seed when generating your DOEs, say 12345 or something. That way, if you ever needed to create the same DOE table again, you re-use the random seed when you generated it, and you'll end up with the same DOE the next time around. You can set this in your DOE dialogue window by clicking the red hot button and then selecting Set Random Seed.

3. If you have the DOE data table already saved, then just open it back up and enter the data.

  Why is it not possible to import data from Excel? This is normally pretty easy with either the JMP Excel add-in that converts Excel data tables to JMP, or through JMP, you can open Excel files and assign columns and rows to import as needed.

  Once you have the data back in JMP, you should be able to go to DOE > Definitive Screening > Fit Definitive Screening to begin your model fit.

Hope this helps,

DS

[#6]

Hallo,

Ok, thanks for the hint concerning the addin in Excel. I did not use it yet and forgot about it.

Of course I can open the data file and add the results.

And yes, I see I can use the given option to model it.

However, I would have liked the spontaneous evaluation as present in the moment

where I entered all data. Is this possible?

It seems to be very tedious to create the model by the given option.

Sincerely

Thommy7571

PS: Words on the GUI are not always clear in German (sometimes not in English):

"Anpassen": means "to fit"   but could also interpreted by "to modify" "Model anpassen" would be clearer.

[#7]

Hi @DrThWillms,

There are many ways to create a model in JMP, especially if you are using a Definitive Screening design.

You can use The Fit Definitive Screening Platform, The Fit Two Level Screening Platform (even if you have 3 levels, it can include quadratic effects in the effects testing), Fit Least Squares (JMP) and Generalized Regression Models (JMP Pro) throught the Fit Model platform, ... not to mention other modeling strategies like Machine Learning and the associated algorithms.

I don't understand your point :

---

@DrThWillms wrote:

However, I would have liked the spontaneous evaluation as present in the moment

where I entered all data. Is this possible?

It seems to be very tedious to create the model by the given option.

---

What are your difficulties ?

Is it a problem of defining the effects to be tested ? The Fit Definitive Screening lacks the flexibility of the Fit Model platform where you can specify the terms to be tested in the model. It will by default test all main effects, 2-factors interactions, and quadratic effects, following a sequential approach : it first select and fit main effects, and in a second step it will select and fit second order effects (2-factors interactions and quadratic effects) based on effect heredity. More infos about this 2-stage approach here : Statistical Details for the Fit Definitive Screening Platform

I would recommend try fitting models in different ways/ with different platforms, to better understand where they agree and disagree. Then, based on statistical evaluation and domain expertise, you can select and/or combine the most relevant model(s) and run validation points to confirm the relevance and accuracy of your model.

Hope this answer may help you or contribute to the discussion,