Confidence intervals on profiler using expanded formula

[#1]

I used the Fit Curve platform to fit Gaussian curves to some viscosity vs salt% data obtained for a range of different formulations. I saved the parameter estimates for Peak Value, Critical Point and Growth Rate to my data table. Following this, I used linear regression to predict the three Gaussian peak parameters using formulation composition as predictors. I saved the prediction formulae and SE for each of the Gaussian parameters to my data table. I manually created a new formula column that calculates predicted viscosity using the Gaussian parameters predicted by the regression model. I then used to profiler with expanded formula box ticked to visualise how the viscosity curves change with formulation composition. The issue is that the profiler doesn't show the confidence interval for the predicted viscosity data. Is there any way to use the SE for the predicted peak parameters to get the confidence interval in the profiler?

[#2]

Hi @CentroidError56 ,

If you can provide the data table that would be helpful. There is a trick to get the SE included in the profiler, you need to have your prediction formula columns named as "Pred Formula [Name of Variable]" and your SE column named as "Pred SE [Name of Variable" - so for example "Pred Formula Growth Rate" and "Pred SE Growth Rate". Then take both columns to Graph>Profiler and JMP will ask you if you want to use the column as the SE.

Thanks,

Ben

[#3]

Hi Ben,

Appreciate your response.

The columns are named correctly. I'm not actually profiling Growth Rate, I'm profiling predicted viscosity which uses Pred Formula Growth Rate as a variable;

The problem is that the SE associated with the Pred Formulas doesn't propagate through to predicted viscosity. Therefore I get no CI when I profile it.

[#4]

Hi @CentroidError56 ,

So when you open the Profiler window on Graph>Profiler are you adding in the 'Whole' Prediction Formula (as you have above) and the PredSE columns?

Thanks,

Ben

[#5]

Hi Ben,

Thanks again for the assistance.

Yes, I have tried entering the Whole Prediction Formula along with the three PredSE columns. This leads to PredSE being additional Y variables in the profiler, rather than being used to calculate the CI for viscosity.

Thanks,

Mark

[#6]

Hi @CentroidError56 ,

If you can send a dataset over I can have a look and have a mess around to see if there's a solution.

Thanks,

Ben