For power studies beyond the simple 2 sample t-test it is probably best to use simulations to estimate the power.
This Answer gives more detail and an example (though not of weighted regression, but the idea would be similar).
Greg Snow
- 52.9k
- 2
- 112
- 190