281085_en-US

liliconcom

Doe

I have six factors. First four are discrete X1, X2, X3 and X4 ( three are 2 level and one is 3 level). I would like to use only one factor a time from first four ( either one from X1, X2, X3 and X4) with remaining two factors. 

for example, I want to use only one kind of sugar at a time.

X1: Regular sugar, 1 or 2 grams

X2: Brown sugar, 1 or 2 grams

X3: Honey, 1 or 2 grams

X4: Sweetener, 1, 1.5 or 2 grams

Thanks

 

Advanced Statistical ModelingQuality and Process EngineeringRe: Doe

A test does not require either a design or a model. Simply test the conditions of interest and select the condition that produces the best result based on your goals.

Re: Doe

I think you are correct it is a test for now.

honestly I don’t know how to use different models based on test or experiment 

 

any comments/suggestions on it ?

 

thanks

Re: Doe

This situation sounds like a 'test,' not an 'experiment.' A test sets up a condition and observes the response. You test different conditions until you obtain the result that you want. You 'pick the winner.' The best choice is always one of the conditions in the test. An experiment sets up conditions that are optimal for fitting a model. The model is used to 'predict the winner.' The best condition might not be one of the conditions in the experiment.

 

Which mindset are you using? Both mindsets are valid, but they are not at all the same.

Re: Doe

Thanks for the answer.

Regarding 1.5 mg only for one ingredient, this experiment is for a drug formulation and I have a doubt that this particular ingredient will not be allowed for clinical trials at higher concentrations.

 Big thanks again 

Re: Doe

View more...

Thanks Mark for this detailed step by step answer, that is what I was looking.

Regarding 1.5 mg only for one ingredient, this experiment is for a drug formulation and I have a doubt that this particular ingredient will not be allowed for clinical trials at higher concentrations.

 Big thanks again 

Re: Doe

If you want the levels for sweetener to be different than the other variables, you can nest the amounts inside sweetener.  Realize when you nest the levels inside sweetener, you will give upon the ability to assess interaction effects with that variable.  As @Mark_Bailey questions, why would you want to do that?

Re: Doe

I recommend using DOE > Custom Design for this purpose. It produces an optimal design based on your specifications.

 

1. Select DOE > Custom Design
2. Double-click Y and enter the name of your response.
3. Change the Goal to match the purpose of your response.
4. Click Add Factor and select Categorical > 4 Levels.
5. Type Sweetener for the name, then press Tab to enter the four choices for sweetener.
6. Click Add Factor and select Continuous.
7. Type Amount and enter 1 for Low and 2 for High.
8. Click Continue.
9. Click Interactions and select 2nd if you think that the effect of changing the Amount is not the same for all choices of the sweetener.
   
   
   
   
10. Enter the number of runs that you want. The minimum number is equal to the number of parameters in the model. I recommend running more than the minimum. I will use 16 but the actual number depends on several aspects of your situation.
11. Click Make Design.
    
    
    
    
12. Click Make Table when the design is ready.

 

Please see Help > JMP Documentation Library > Design of Experiments guide for more information and examples.

 

Why do your want runs with Amount = 1.5?

Re: Doe

Thanks Mark for your quick response,

 

I am new to JMP, can you show me how to join categorical factors and continuous factor.

for example, sweetener to be only 1, 1.5 or 2 grams.

 

Thanks

Re: Doe

I suggest creating one categorical factor to identify the Type of sweetener (i.e., sugar, brown sugar, honey, and sweetener) and a continuous factor Amount (i.e., 1 or 2).