CausalQueries
CausalQueries copied to clipboard
integrated-inferences
parmap bug (os dependent?)
test_model <- make_model("X -> M -> Y <- X; M <-> Y") test_parmap <- get_parmap(test_model)
Returns on my computer the weird parmap attached:
So what should be data types are also things like X0M0Y1.1 Remember thinking when first seeing this that it's got something to do with the model being confounded, but checked this once with Macartan and he was not having the problem, i.e. he only had data types n the column titles. I would mention I'm on macOS, the problem was not there on windows back in the day, but not sure if this is necessarily true.
Would mention that I recall seeing stuff like this happen with the column titles when using code such as:
df <- data.frame( X = c(..), Y = c(..), )
But I cannot remember the exact circumstances, will see if I can get that happening again. Any ideas? @till-tietz
Giving this literally one more second of thought would have made me realize that this obviously happens because a column is defined twice, see: df <- data.frame( x.0 = c(1, 1), x.1 = c(2, 3), x.1 = c(2, 4) )
Generally not the only way wrong column names can happen, but rather obvious in our case. The weird part is that the doubles are not consistent with the same parameters
What version are you on? Can't reproduce this on the 1.0.1 Linux binaries either.
updated cq 2 weeks ago I think. But I have had this issue in all versions of the package!
This is very strange. Looks like we'll have to check the mac OS binary builds since this issue doesn't seem to reproduce on 1.0.1 for windows + Linux
Taking the OS dependent stuff back: this is happening on the windows PCs at WZB as well, will send version of cq there soon
@cristian-nicolescu any update on the version that produces the bug? There have been recent updates to parmap that are now on main. Would be curious if those resolve the issue
Very sorry for the late reply: On mac both cran and dev the (newest) versions still have the bug, will update on windows asap