Time dependent propensity models

[marycombs]

Time dependent propensity models

[osofr]

Not sure, can you provide an example? Perhaps the easiest is to use time-dependent interactions, but it's not trivial.

…

On Wed, Jun 26, 2019, 10:31 AM marycombs ***@***.***> wrote: Different from stratified ps models: how do I specify different variable sets in g_form_TRT? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#44>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABTI6TQYHEXL35E455EJ3KTP4ORWXANCNFSM4H3UPZPQ> .

[marycombs]

For example, if I want to model probability of treatment at time t on “entire treatment history” and baseline covariates, I.e. in an SCM, A(t) = f_{A(t)}(W, \overline{A}(t-1)).

In order to include treatment at previous time points I would create lagged variables and specify them in gform_TRT, however, even if I stratify on time, the only rows with no NAs in the lagged treatment variables will be the last row of an individual with the longest observation time.

Thank you!

[marycombs]

Is there any followup regarding this? Is my question clear?