Hello everyone,
as a long-time reader of Statalist the time has come to pose a question of my own. I am trying the estimate the effect of a continuous treatment x on an ordinal outcome y (i.e. has the perceived risk of job loss gone up/down/remained the same). To identify the causal impact of x on y I have a set of (weak) instrumental variables, so I was trying to find an IV version of Stata's oprobit command. I came across Stata's eoprobit - the documentation explains how eoprobit lets you specify a model for endogenous covariates but never directly mentions instruments or the standard first stage model statistics. Therefore, I wanted to ask two clarification questions:
1. Is eoprobit essentially the IV version of oprobit or am I missing anything?
2. If eoprobit is in fact equivalent to oprobit with IV, is there any way to conduct weak instrument robust inference as with ivreg2?
Thanks for your help!
as a long-time reader of Statalist the time has come to pose a question of my own. I am trying the estimate the effect of a continuous treatment x on an ordinal outcome y (i.e. has the perceived risk of job loss gone up/down/remained the same). To identify the causal impact of x on y I have a set of (weak) instrumental variables, so I was trying to find an IV version of Stata's oprobit command. I came across Stata's eoprobit - the documentation explains how eoprobit lets you specify a model for endogenous covariates but never directly mentions instruments or the standard first stage model statistics. Therefore, I wanted to ask two clarification questions:
1. Is eoprobit essentially the IV version of oprobit or am I missing anything?
2. If eoprobit is in fact equivalent to oprobit with IV, is there any way to conduct weak instrument robust inference as with ivreg2?
Thanks for your help!
