Alter the blavaan source

[RensDofferhoff]

Was there any reason not to use blavaan from cran? @maltelueken

[RensDofferhoff]

The unit test numbers are ever so slightly different.
Slight difference in sampling params?
Or was this the reason to use a clone of blavaan?

[maltelueken]

The reason why we use a clone of blavaan is that after version 0.5-2, a bug was introduced which I could not work around. I saw that they just released a new version 0.5-5, but this seems to introduce new bugs (at least in our module). Previously, it wasn't possible to pin R packages to a specific version, so I used a clone. Is this possible in JASP now?

[maltelueken]

I am not sure why there are these slight differences in the parameter estimates that make the Bayesian tests fail. They run fine locally. It's likely due to the sampling in Stan indeed. There is a similar issue for the classical tests on Ubuntu, also very small differences.

[RensDofferhoff]

Previously, it wasn't possible to pin R packages to a specific version, so I used a clone. Is this possible in JASP now?

I have finished a renv lockfile feature started by @vandenman that allows this.
The only known remaining issue is blavaan compilation on MacOS.
Therefore I had hoped to pin it to precompiled CRAN binaries.

But alas. I will continue to fix the blavaan compilation.

[maltelueken]

Great! Is it not possible to pin it to 0.5-2 on CRAN then?

[RensDofferhoff]

Yes, once/if its merged!

[RensDofferhoff]

Though CRAN no longer has those binaries

[maltelueken]

Ok, but now it still pins to version 0.5-5. I am a bit confused.

[RensDofferhoff]

CRAN only keeps the binaries of the latest version.
I will remove this PR for now since we need 0.5-2 due to bugs in later versions.
We can pin it to 0.5-2 once jasp-stats/jasp-desktop#5556 is merged