How to use prediction profiler to achieve a combination of target conditions?

[#1]

I have a 5 column data. Variation of input parameters/conditions in rows of the first three column results in (measured) values in column 4 and 5. 

How to use prediction profiler to predict a combination of input parameters/conditions to achieve a target combination (in my case a target concentration and depth)?

I would prefer a graphical way but I can also handle JSL.

[#2]

You must first fit a model to the data. Here is the most general solution. Short-cuts are possible in some model cases. Fit a separate model to each response. Save the model for each response. Select Graph > Profiler and add the two saved formula columns to be predicted.

Unfortunately, three observations limits the modeling that is possible. Here is a linear regression model without an intercept.

[#3]

Thanks. When you say "fit model" , should this be a known model for the underlying mechanism which affects the output parameters for given input parameters? 

I need to understand these two steps better

1. Fit a separate model to each response.

2. Save the model for each response. 

Could I please get a breakdown of these two steps?

[#4]

That's right. It might be a theoretical model. It might be an empirical model, though. You are limited to estimating no more than three parameters. For example, I should how to profile the response using a linear regression with three parameters, one for each factor, but omitting a constant term (intercept). This example is an empirical model.

There might be additional effects of any of these factor such as interaction effects and non-linear effects, but three observations are insufficient to model or explore such effects.

You might consider using a designed experiment to collect the data for estimating the desired parameters and include extra runs for error degrees of freedom. You could start with the three runs that you have in hand and augment them in an optimal way. My suggestion assumes that you can control factor levels. That is to say, your data is not an observational study.

As to how to follow my suggestions, please see the JMP Help for fitting linear models, in general, and launching the Fit Least Squares platform, in particular, if you want to replicate my result.

[#5]

hello guys, in the prediction profiler, for example in the flow where you can adjust it. is there a way in a kategorical variable to include them?

as you can see in the category coal, is there a way to find the interval to reliance by selecting all of the coal variables? so i dont only find the answer when i am looking at coal but also natural gas and oil?

[#6]

The reliance depends on the type of coal. There might also be interactions between Coal and the other two variables that would add additional dependence on Coal. You would want to find the optimum for each coal I would think.