wilcas / vaecit

Variational autoencoders to use latent factors in causal inference
Apache License 2.0
0 stars 0 forks source link

Figure out why independent is being called as causal mediation #13

Closed wilcas closed 5 years ago

wilcas commented 5 years ago

if you see the simulations I have thus far, it seems that independent association is getting called as full mediation when it definitely shouldn't be.

At first it seemed like the reason may have been that I was using more genotypes, but the issue could be:

  1. Having really large integer effect sizes, and
  2. Having fixed numbers anywhere as coefficients