Scale of weights in brms leads to regression discrepancies?


Given a very simple example:

fit <- brm(y|weights(z)~x, data=df)

If I use z as weights, or z*1000 by weights, not only do the uncertainties changed (I understand why this would be the case, since brms interprets weights as literal number of observations), but why does my regression line change as well?

Is there an appropriate scale for weighting? If I use the 0-1 scale the uncertainty is incredibly large…


Please also provide the following information in addition to your question:

  • Operating System: Windows10
  • brms Version: 2.10

You can scale the weight if needed. See ?addition-terms:
scale Logical; Indicates whether weights should be scaled so that the average weight equals one. Defaults to FALSE.

Could that be the problem?