Hi All;

This is my first time using DOE tool in JMP and I have some issues trying to understand the results that I got by applying different analysis for the same set of data.

I am trying to find the combinations that minimize the amount of rejects I got when apllied different heat sources around the assembly process, in order to determine the best way to reduce piroelectric damage in our components.

DOE consists of 12 runs were the following variables were evaluated:
FILTER: Yes/No
CURE TIME: 30 min/ 10 min
CURE OVEN: Pyro Sensitive 150 degrees/Pyro Sensitive 175 degrees / No Pyro Sensitive

When I used the Custom Design tool in JMP if i applied the Main Effect I got that in order to minimize the rejects I need to Use Filter/30min/Pyro Sensitive 175.

However If I make the same analysis but at this time I didn't click the option of Main effect I got that in order to minimize the effects I must use No Filter/30min/Pyro Sensitive 175.:confused:

Does this mean that filter is not significant?...I mean it is the same use it or not...

:thanx:

I asked one of our JMP experts to assist. He is located in the Philippines, so you may not receive an immediate response.

Could you provide some background information on your design?



What makes this a custom design? Was it not a standard classical design or an Orthogonal Array?
Were there any restrictions in randomization when the experiment was run? Such as a difficult to change characteristic that was only changed one time?
What does the ANOVA table indicate? That is, what was the p-value for filter?
Your post seems to indicate that the response was count or proportion data. You should transform this type of data using the Freeman Tukey transform before analysis. I am a Minitab user, so I do not know what JMP transforms are available. The formulae are shown in this link (http://www.spatialanalysisonline.com/output/html/Datatransformsandbacktransforms.html). The first FT transform is for count data, the second FT transform is for proportion data.

Well as you can see I'm not an expert on this topic...I used a Full factorial 2x2x3.

There wree some restrictions in randomization since the Dicing Saw process is first and we use filter and different Cure Time on this process before go to Mount Operation were we used the Cure Oven so I took 4 different wafers each wafer divided in three diferent groups one each cure oven so I used the table that JMP brings me to perform the DOE.

Dicing Saw Filter Heating Cure Time Mount Cure Oven
YES 30 Pyro Sensitive Profile 150
YES 30 Pyro Sensitive Profile 175
YES 30 No Pyro Sensitive
YES 10 Pyro Sensitive Profile 150
YES 10 Pyro Sensitive Profile 175
YES 10 No Pyro Sensitive
NO 30 Pyro Sensitive Profile 150
NO 30 Pyro Sensitive Profile 175
NO 30 No Pyro Sensitive
NO 10 Pyro Sensitive Profile 150
NO 10 Pyro Sensitive Profile 175
NO 10 No Pyro Sensitive

By looking at each of the responses I was evaluating I got the results attached.

Hope this helps.

Did you run the experiment in the specific order shown above? Restricted randomization introduces complications in the analysis. Such experiments are called split plot designs.

The ANOVA table is labeled Effects Tests in your report. The p-value is labeled Prob > F. Normally a p-value </= 0.05 is used to determine significance. A p-value between 0.05 and 0.10 usually warrants additional investigation.

Hi All;
When I used the Custom Design tool in JMP if i applied the Main Effect I got that in order to minimize the rejects I need to Use Filter/30min/Pyro Sensitive 175.

However If I make the same analysis but at this time I didn't click the option of Main effect I got that in order to minimize the effects I must use No Filter/30min/Pyro Sensitive 175.:confused:

Does this mean that filter is not significant?...I mean it is the same use it or not...
:thanx:

Hi Raquel.
Welcome to the Cove!.
Let me address your question but first thing first. You have two different models (clicked on main effect vs did not clicked on main effect) that gave two different optimum settings (Use Filter/30min/Pyro Sensitive 175 vs. No Filter/30min/Pyro Sensitive 175). Then since thier recommendations are different you are asking if the filter is significant or not.
The analysis flow should be reversed. First create a model which includes only the significant factors. Then if the model is already satisfactory you use it to find the optimum setting.

Im not sure what you mean by the main effect option, but if you are trying to do a full factorial, here is the step by step way to do it in JMP.

Thanks for all your help Reynald ...:applause:

I did it Steb by Step what you mentioned but based on the file, and it loooks none of the factors are significant.

However when I aplied the same but using the Custom Design and clicking at the Main Effect button I got the results I attached in the second tab of the second file. The first tab of the second file is the Effect Leverage analysis of the Full Factorial Design.

I show you what button I am clicking on the JMP DOE presentation you sent me.

Thanks for your help again.

:confused:

Hi,
Now I understand.
That button "Model" would add the main effects terms in your model.
You dont notice what happens however because by default they are already included. Try to remove the terms and then click on this button and you will see it.
11309

Then about the difference in the results, the real cause actually is that by using the full factorial design, it includes all 1st level interaction terms. On the other hand building the model using custom design by default includes only the main effects term (unless you add the interactions). You will notice that if your compare these results:
11310

Hope that helps.