Open haziqj opened 1 month ago
Hi @haziqj! No worries. Thank you for the update, I'll take a look at Yves' script.
Thus, the idea is to let the C++ code be fully contained in {plFA}
, which should export some wrapper functions directly callable in {lavaan}
. Did I understand correctly?
Sure, no rush on my side. It's a busy period for me, too, but I'll start looking into it.
Best, Giuseppe
Thus, the idea is to let the C++ code be fully contained in {plFA}, which should export some wrapper functions directly callable in {lavaan}. Did I understand correctly?
Yes, correct. In a sense we are creating a plugin for lavaan. At the top of my head, it could be as follows:
Ok, I see. It sounds good.
Surely it requires some tinkering on the {plFA}
side, but it should be feasible.
I'll pick it up in the next weeks.
Have a nice weekend :)
Hi @giuseppealfonzetti. Sorry for the radio silence. I am currently travelling so not able to look at our project just yet.
However I wanted to share with you some broad ideas on what needs to be done. To understand better I share with you this R script by Yves: https://github.com/yrosseel/lavaan/blob/master/inst/understanding_lavaan_internals.R It's a little outdated but in principle I think it's still the same. When lavaan is called it goes through a series of steps, and the steps are often conditional on the estimation method.
Our job essentially is to create a new
estimator = "SPML"
(which stands for stochastic pairwise max likelihood), and then replace the necessary functions with functions you created in the{plFA}
package.Importantly Yves requests that
{lavaan}
contain no C/C++ code (at least for now), which means that the{plFA}
will have to coexist with{lavaan}
. Might require some tinkering at the plFA side to get things working smoothly too.Anyway I will pick this up closer to July, when I can afford to dedicate time to this. If you have any questions in the mean time feel free to ask.
Haziq