More than a GUI for specifying the model, I'd value a graphic output: A drawn path diagram, both of the model during construction (so people can see what it is they have asked for), and a publication-ready path output in svg or pdf of the run model with loadings etc.
It is difficult to have publication quality graphics emerge automatically. Often the user needs to modify the figure after it has been produced, and depending on how the image has been constructed, it isn't always easy, especially multiplatform.
A limitation of generating graphical output is that it will work for pathic but not mathic input scripts. Several other pieces are missing too; we have no obvious graphical representations of constraints, nor thresholds, nor mixture distributions.
Sadly, the convention of explicitly showing the variance (self-covariance, possibly residual) of a variable has not caught on. Path diagrams lacking this necessary information are found all the time. Usually it's pretty obvious that all latent exogenous variables are standardized, although in some cases it isn't and it can be hard to recover exactly what an ambiguous diagram was supposed to specify. Any thoughts on how to improve this situation?
GUI: path output
More than a GUI for specifying the model, I'd value a graphic output: A drawn path diagram, both of the model during construction (so people can see what it is they have asked for), and a publication-ready path output in svg or pdf of the run model with loadings etc.
Log in or register to post comments
In reply to GUI: path output by tbates
GUI:path output
I agree.
Log in or register to post comments
In reply to GUI:path output by smcquillin
GUI
It is difficult to have publication quality graphics emerge automatically. Often the user needs to modify the figure after it has been produced, and depending on how the image has been constructed, it isn't always easy, especially multiplatform.
A limitation of generating graphical output is that it will work for pathic but not mathic input scripts. Several other pieces are missing too; we have no obvious graphical representations of constraints, nor thresholds, nor mixture distributions.
Sadly, the convention of explicitly showing the variance (self-covariance, possibly residual) of a variable has not caught on. Path diagrams lacking this necessary information are found all the time. Usually it's pretty obvious that all latent exogenous variables are standardized, although in some cases it isn't and it can be hard to recover exactly what an ambiguous diagram was supposed to specify. Any thoughts on how to improve this situation?
Log in or register to post comments
Categorical data thresholds implemented
Wish #1 granted: Thresholds for binary and ordinal, as well as mixed continuous and ordinal data are fully implemented.
Log in or register to post comments