Hello,
I am new to DoE and JMP and I am already facing some problems in Custom Design.
I have already designed the experiment, meaning, I have alredy clicked "Make a Design" and "Make a Table", when I wanted to change one of the factors value (value range). After changing the values, I clicked "Make a Design" and "Make a Table". Then a new window opened, with new values, but table in the background stayed the same. So I repeated clicking "Make a Design" and "Make a Table" several times.
What it seemed to happen was, that sometimes the new table was completely diffrent to primary table and sometimes everything stayed the same, except for the values I have changed (and also those values have changed equivalently - the highest values for the new highest values and so on).
My question is: what is the proper way to change value of factor once the table is already made and what does JMP do?
I would be really grateful, if anyone could answer
Kind regards
Andreja
First of all, you have not done anything wrong! JMP behaves this way because experimenters often consider more than one candidate design before selecting the final design.
As to the differences between the results, Custom Design is based on an optimization algorithm that starts with a horrible random initial design and then iteratively improves it. There are cases where the nature of the definitions of your factors, models, and so on might have more than one solution or optimal design. When you click Back, you are telling JMP that you don't want this design and try to make another. Only if there is just one solution will you get the same result every time.
If you are only changing factor levels, then close the latest data table, click Back, edit the factor levels, and click Make Design again.
If you already ran the experiment but were unable to achieve the desired factor levels, then edit the data table to reflect the actual levels for each run.
First of all, you have not done anything wrong! JMP behaves this way because experimenters often consider more than one candidate design before selecting the final design.
As to the differences between the results, Custom Design is based on an optimization algorithm that starts with a horrible random initial design and then iteratively improves it. There are cases where the nature of the definitions of your factors, models, and so on might have more than one solution or optimal design. When you click Back, you are telling JMP that you don't want this design and try to make another. Only if there is just one solution will you get the same result every time.
If you are only changing factor levels, then close the latest data table, click Back, edit the factor levels, and click Make Design again.
If you already ran the experiment but were unable to achieve the desired factor levels, then edit the data table to reflect the actual levels for each run.
I have run an experiment already, but with different boundary values.
If I change these values in a table, the prediction profiler shows the result as if I had old boundaries.
Please let me know how to solve it?
Try editing the Coding column property. I suspect that the old limits are still defined by that property. Change them to reflect the new limits.
Hi Mark,
If some factor levels are not obtainable due to lets say 'tool limitations', and I need to use slightly different factor values for a few trials.
If I need to edit these values right in the table, will my analysis (Fit Model) run based on the updated values, or the ones I originally entered?
Some statistical error would be inherent in the design if the code behind the screen doesn't account for the 'real' values used during the experiment vs what it originally wanted me to use.
Thanks!
-Charlie
JMP will perform the analysis based on the current values in the data table. Your updated values will be used in the regression analysis.