Is it possible easily to add Predicted and Residual values to the table input to Structural Equation Modeling; and, to be shown the Predicting Equation, please?

[#1]

*

Subject: Is it possible to add Predicted and Residual values to the table input to Structural Equation Modeling; and, to be shown the Predicting Equation, please?

I have had some good success learning to use the !neat! Structural Equation Modeling module demonstrated by @LauraCS. But I have not yet been able to figure out how to add Predicted and Residual values to the table used in modeling. I have also not yet figured out how readily to capture the predicting equation. Nor do I see a new variable name created for the Predicted values. Although I am able to formulate a prediction equation given the tabulated coefficients for each variable, and to calculate from that the residual values, I am hoping that those features exist and that I simply can't find out how to invoke them.

I see Residual Plots and Conditional Residual Plots mentioned at https://www.jmp.com/support/help/en/15.1/index.shtml#search/residuals | Structural Equation Model Platform | Model Options, and the nine hits there for "residuals"; but none of that seems to address my need for a way simply to add Predicted and Residual values to the open table and to be able to capture the Predicted Equation without having painstakingly to reconstruct it from coefficients. 

Either I'm missing something [please tell me what ;-)] or maybe my needs should be added to a wish list.

Thanks for any guidance you can offer on this.

Sincerely, @gambist0 

[#2 · Accepted solution]

Hello @gambist0,

Thank you for your question! The most common way to look at residuals in SEM is through the "Normalized Residuals" and/or "Normalized Residuals Heat Map" option in the red triangle menu of a fitted model. Because the data that are being modeled are the variances and covariances among the variables, the residuals are based on those data, which as you point out, are not the individual observations in the data table. Version 15 doesn't have the feature you request. The good news is that we're currently working on a number of features that will address your needs, including a profiler so you can look at predictions for individual equations in the model. This profiler is already available in our Early Adopter program. Moreover, we're actively working on saving the observational residuals to the data table.

Best,

~Laura

[#3]

Thank you, @LauraCS. This has been an interesting side trip. All best wishes. @gambist0

[#4 · Accepted solution]

Saving normalized residuals in a  separate JMP table will be a great option to add in version 15.3. For now, we can right-click and save normalized residuals as data table, then use Stack to convert those multiple columns into a single column and then do residuals analysis via describe menu on that single column to check for normality etc.