Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: xthtaylor by hand or using xtivreg2

From   Nick Cox <>
Subject   Re: st: xthtaylor by hand or using xtivreg2
Date   Wed, 2 Nov 2011 09:30:50 +0000

Thanks. That adds more detail. It's over to people who do this kind of
thing to see if they can say anything helpful.

On Wed, Nov 2, 2011 at 9:25 AM, M.M. Kramer <> wrote:
> Dear Nick,
> I am trying to estimate the impact of a self-choosen treatment on portfolio
> returns of retail investors. This self-choosen treatment is very likely
> endogenous, it does not vary over time and we do not have external
> instuments available, so Hausman-Taylor might be an appropriate estimation
> technique.  The problem that I keep encountering is that I cannot assess the
> quality of the instruments that xthtaylor creates. I cannot see the first
> stage results when using xtoverid2 and I receive warnings on collinearity.
> My idea was to create the instruments by hand and then use 2SLS to see
> what's going on. So far, I found that xthtaylor transforms the timevarying
> exogenous variables into means, and demeaned variables and uses some other
> transformations. My question therefore is how exactly to replicate the
> variable transformation of xthaylor to create the instruments by hand.
> Marc
> Nick Cox wrote:
>> Many people will sympathise to some degree here, but what kind of help
>> are you expecting _which can reasonably be provided_?
>> 1. To comment helpfully on your difficulties with -xthtaylor- people
>> will surely want more detail on your problems.
>> 2. If you seek to use -xtivreg2- (SSC), the same comment applies
>> there: the help is the place to start and people will want to see
>> specific queries.
>> 3. -xthtaylor- is some hundreds of lines long, so reproducing it "by
>> hand" is not trivial.
>> Nick
>> On Wed, Nov 2, 2011 at 8:26 AM, M.M. Kramer <> wrote:
>>> I have been struggling for some time with the xthtaylor command.
>>> Especially
>>> the various error messages that occur after some tests are hard to solve.
>>>  Does anyone have any clear instructions on how to transform the
>>> variables
>>> in a way that xtivreg2 can be used?  Or may be anyone has the syntax to
>>> replicate the output of xthtaylor by hand.
>>> Thanks.

*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index