falkcarl / ockhamsem Goto Github PK

Looks great! Only comment is to change lower.tail option for CFI for the plot, and maybe summary.

Might make fitting same model to many different correlation matrices easier, if that's supported. Thanks to Terrence Jorgensen for the suggestion.

summary.fitprop generates warnings from ks.test about p-value not being accurate in presence of ties. Clean up or do something with this.

e.g., one that can be accessed via ?ockhamSEM

(thanks Kris Preacher)

Applies to matrices returned from clusterGeneration. Some are apparently not symmetric at some very small decimal place. Currently rounded to 5 decimal places.

We should write some unit tests to make package maintenance easier.

Probably some scattered throughout fitprop.R

Affects summary.fitprop so far. Does lower.tail not change the sign of the effect sizes?

Not sure if we need it yet. There's not that much code nor that many functions.

Applies to Onion and other clusterGeneration methods. Currently a bit ad-hoc.

Does it really need to be loaded again in order to make a cluster?

Checking package leads to lots of notes about functions from other packages that are called, and something about globally defined variables

Hi,

Thank you for this interesting project and providing such a helpful R package!

When looking at the source code, I noticed that you compute the NML as -2*Log-Normalized Maximum Likelihood. I am just curious: What is the rationale for computing -2 times the log-NML? Thanks!

Some versions of Rcpp generate an error when trying Euler plots, similar to that here: https://stackoverflow.com/questions/68416435/rcpp-package-doesnt-include-rcpp-precious-remove

Has code from Harry Joe's group that includes onion, vine, and more. We could just import relevant functions and use those instead of having our own copy of onion and c-vine methods.

Update DESCRIPTION to include any credit for others' code.

Different ranges supported (defaults to 0 to 1), but we don't have much for illustration. Add to docs or some FAQ. logl is an example.
(thanks Kris Preacher)

For any other functions we want to be readable outside of the package. May include fiddling with how S3 objects are created and S3 relevant functions are exported: http://r-pkgs.had.co.nz/namespace.html

Consider adding NML and the "uniform index-of-fit" (i*) described by Botha, Shapiro, and Steiger (1988)
(thanks again to Kris Preacher)