'weights' vs 'method.rsp = "p.weight"'

[MartinOFlahertyAus]

Hi - I wanted to ask what the difference is between using IPTW weights in the 'weights' option vs specifying the method.rsp = "p.weight" option. My basic understanding is that:

• specifying 'weights' will alter the contribution of each observation to the likelihood during the model fitting process (so the underlying function estimated ), and the individual treatment effects (y(1)i - y(0)i) will be weighted in calculating the average treatment effect.
• specifying 'method.rsp = "p.weight"' will fit the same underlying function as an unweighted model, and only apply the weights in calculating the average treatment effect.

Is that correct? Do you know of any evidence about which approach performs better?