I am not Stephen Jenkins and my own practical experience is with
right-censored data, which is obviously much left problematic.
But in your case, and wearing my hat as a non-statistical-specialist
reviewer of your paper, I would be concerned that those companies which
have websites so early in the Internet era, and which you will probably
have to drop from your analysis as others have said, are systematically
different from latecomers to the Age of the Web. I would very much want
to know if there are statistically significant differences between the
dropped group of companies and the non-dropped group of companies in
terms of your key explanatory variables. So beyond pointing out that
the results don't really differ, it would be even more convincing if you
could say that dropped companies look a lot like non-dropped companies.
This is probably an obvious point, but you don't mention it, so maybe
you haven't done this sort of poking around.
-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Daniel Simon
Sent: Thursday, November 16, 2006 10:15 AM
To: statalist@hsphsun2.harvard.edu; statalist@hsphsun2.harvard.edu
Subject: Re: st: left censoring in discrete-time duration model
Austin - thanks for the response. In fact, I have done what you
propose.
My point (3) where I said "I drop all observations from 1996 for firms
that
had websites" is actually what you are proposing, because once a firm
has
established a website (transitioned), they are excluded from the
analysis
sample thereafter. I did not think about that when I wrote the
sentence.
My apologies for not being clear. Thanks again for your response, which
reminded me of what I was actually doing.
Moreover, as Austin says, I would also like to hear whether others also
this is a sensible approach. thanks.
Daniel
At 12:00 PM 11/16/2006 -0500, Austin Nichols wrote:
>Daniel Simon--
>It seems to me that your estimates can only apply to introductions
>("failures") after 1996, since you cannot distinguish pre-1996 and
>1996 introductions, and you should drop firms (in all years) that had
>websites in 1996, while keeping data on all other firms from 1996,
>though I would be interested to hear from someone who actually runs
>these sorts of models, e.g. Stephen Jenkins.
>
>On 11/16/06, Daniel Simon <dhs29@cornell.edu> wrote:
>>Hi - I'm using -hshaz- to estimate a discrete-time hazard model. I
have
>>some left censoring that I'm not sure how to deal with. I am looking
at
>>firms establishing websites. I can only observe the introduction of
>>websites from 1996 onwards. However, I know that some firms
established
>>websites prior to 1996, but I'm not sure which ones. Currently, I have
>>tried three approaches: (1) Treat all firms that had a website in 1996
as
>>if they adopted in 1996 (the first year of the sample period), whether
they
>>adopted in 1996 or adopted earlier; (2) Exclude 1996 from the sample
(begin
>>the analysis with 1997); (3) Drop all observations from 1996 for firms
that
>>had websites.
>>All three approaches give me quite similar results, so it does not
appear
>>that the censoring is a major issue. But, I'm wondering if there is a
>>better way to deal with it. Thanks. Daniel
>*
>* For searches and help try:
>* http://www.stata.com/support/faqs/res/findit.html
>* http://www.stata.com/support/statalist/faq
>* http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/