[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
Re: st: Predicted probabilities after oprobit w/robust standard errors
At 06:55 PM 6/1/2006, Matt Barmack wrote:
Interesting point about the intuition, but it is not right in this
case. The predict command does allow you to also estimate the
standard error of the linear prediction, which I think gets at the
issue you are raising, and the standard error of the prediction is
affected by whether or not you use cluster or robust.
Specifying cluster or robust does not seem to change the predicted
probabilities from oprobit. Does it? Shouldn't it?
Intuitively/naively, I am thinking that for an observation for which
the variance of the random part of the latent index is high, there is
a greater chance of ending up further away from what the
deterministic part of the latent index alone might suggest.
As to why your intuition is wrong - well, if it were correct, it
would mean that you would, say, with higher standard errors, be more
likely to predict that every case was a 5 or a 1 on a 5 pt
scale. Why should that be the case, i.e. why should higher standard
errors make it more likely that cases have extreme values? The
predictions are not affected by the standard errors in this way, but
our confidence in how accurate the predictions are is.
Also, in probit, the residual is assumed to have a normal(0, 1)
distribution. That is true regardless of what the standard errors
are. Together, the linear prediction, the cutpoints, and the
distribution of the residual will determine what the predicted
Richard Williams, Notre Dame Dept of Sociology
OFFICE: (574)631-6668, (574)631-6463
WWW (personal): http://www.nd.edu/~rwilliam
WWW (department): http://www.nd.edu/~soc
* For searches and help try: