I wouldn't characterize your problem as censoring.  You have discontinuous
risk intervals--time periods in which you cannot observe events.  This is
easily handled with a Cox model using start-stop notation (in stata this
would mean specifying when bills enter, exit, and fail when you stset your
data).

For example, if you had a bill that you could observe from day 1 through 7
and then from day 22 on, you would set up your data as:

Bill	Start	Stop	Event
1	1	7	0
1	22	28	1 

If the event was never observed, then you would include the first line only
and it would be treated as a censored observation.  

You can also treat bills left over at the end of the session as censored
(event=0 for last observation).

Tobin


-----Original Message-----
From: Political Methodology Society [mailto:[log in to unmask]] On
Behalf Of [log in to unmask]
Sent: Wednesday, March 28, 2007 12:41 PM
To: [log in to unmask]
Subject: [POLMETH] Unknown type of censoring

Dear all:

I am analyzing the time that legislative proposals spend inside
Parliament.  When a bill is proposed, the time when it is “tabled” is
recorded (along with other covariates).  The bill then proceeds,
unobserved in my dataset, through the meanders of the legislative process.
 When, and if, a bill emerges, the duration of the spell in Parliament is
recorded.

Some bills fail to exit in a typical case of right-censoring: given enough
extra time, they would eventually emerge from the process and join the set
of legislation submitted to the executive for promulgation.

But there are also bills (most of them) that will never emerge: a
committee does not report them, or they fail a final passage vote, or they
are rejected by the second chamber, etc.  None of this is recorded in the
data.  In legislatures that permanently archive all that remains pending
by the time the assembly adjourns, this lack of detail is unproblematic:
all can be considered “dead” by the start of the new Parliament.

But in other parliaments, including the case at hand, all pending matter
is inherited by future parliaments.  I have records of proposals initiated
as long as 11 years before approval, 3 new parliaments afterwards.  And
this poses a problem of censoring that I am unfamiliar with.  A good deal
of the density that remains at risk after the last record is, in fact,
dead matter, which appears to violate the assumption behind duration
models with right-censored data.

Any comment or suggestions will be much appreciated.

Thank you!

--e
-----------------------
Eric Magar
Ciencia Política - ITAM




-- 
La información contenida en este mensaje de datos es confidencial,
constituye un secreto industrial y/o profesional en términos de la
legislación vigente y se encuentra dirigida exclusivamente al
destinatario indicado en dicho mensaje. Si usted recibe esta información
por error o si usted no es el destinatario del mensaje, favor de
notificar al emisor, y destrúyalo.

The information contained in this electronic message is confidential,
it constitutes a professional and/or industrial secret in terms of the
current legislation, and is intended for its recipient only. If you
receive this message by mistake and/or if you are not the recipient
thereof, please notify the sender and destroy it.