st: Re: how to select model in xtpcse , corr(ar1)

classic Classic list List threaded Threaded
5 messages Options
Reply | Threaded
Open this post in threaded view
|

st: Re: how to select model in xtpcse , corr(ar1)

Christopher F Baum
< >
Ghislain said
I want to compare two models one made with : xtpcse x y z,corr(ar1)  
the other with : xtpcse x y w,corr(ar1) xtpcse doesn't return AIC nor  
BIC nor e(ll),

Martin suggests looking at r^2.  AIC and BIC are criteria that will  
penalize for non-parsimonious models. As both of the models you  
specify have exactly the same length, they would not be useful. They  
are not estimated with maximum likelihood, so there is no e(ll). [OLS  
is not estimated with MLE either, but -regress- does provide e(ll)].  
As r^2 is perfectly comparable for models of the same length, r^2  
would seem to make sense as a comparison.

I would worry more about the notion that if each of these models does  
a good job, with a significant z or w respectively, then there is  
surely the risk that they are both misspecified versions of a model  
that contains both. Can you indeed rule out w appearing in the first  
model, or z in the second?


Kit Baum, Boston College Economics and DIW Berlin
http://ideas.repec.org/e/pba1.html
An Introduction to Modern Econometrics Using Stata:
http://www.stata-press.com/books/imeus.html

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
Reply | Threaded
Open this post in threaded view
|

st: left censoring

IOANNA MARIOU STYLIANOU
Hi

iam trying to estimate duration models(Cox-PH) but i have left censoring..Does anyone knows how this is treated in STATA?

Ioanna Stylianou

----- Original Message -----
From: Christopher Baum <[hidden email]>
Date: Thursday, December 4, 2008 1:54 pm
Subject: st: Re: how to select model in xtpcse , corr(ar1)
To: [hidden email]


> < >
>  Ghislain said
>  I want to compare two models one made with : xtpcse x y z,corr(ar1)  
>
>  the other with : xtpcse x y w,corr(ar1) xtpcse doesn't return AIC nor
>  
>  BIC nor e(ll),
>  
>  Martin suggests looking at r^2.  AIC and BIC are criteria that will  
>
>  penalize for non-parsimonious models. As both of the models you  
>  specify have exactly the same length, they would not be useful. They  
>
>  are not estimated with maximum likelihood, so there is no e(ll). [OLS
>  
>  is not estimated with MLE either, but -regress- does provide e(ll)].  
>  
>  As r^2 is perfectly comparable for models of the same length, r^2  
>  would seem to make sense as a comparison.
>  
>  I would worry more about the notion that if each of these models does
>  
>  a good job, with a significant z or w respectively, then there is  
>  surely the risk that they are both misspecified versions of a model  
>
>  that contains both. Can you indeed rule out w appearing in the first  
>
>  model, or z in the second?
>  
>  
>  Kit Baum, Boston College Economics and DIW Berlin
>  http://ideas.repec.org/e/pba1.html
>  An Introduction to Modern Econometrics Using Stata:
>  http://www.stata-press.com/books/imeus.html
>  
>  *
>  *   For searches and help try:
>  *   http://www.stata.com/help.cgi?search
>  *   http://www.stata.com/support/statalist/faq
>  *   http://www.ats.ucla.edu/stat/stata/
>  
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
Reply | Threaded
Open this post in threaded view
|

st: RE: left censoring

Mak, Timothy
Have you checked the -stset- help menu?

Tim

-----Original Message-----
From: [hidden email]
[mailto:[hidden email]] On Behalf Of IOANNA MARIOU
STYLIANOU
Sent: 04 December 2008 20:01
To: [hidden email]
Subject: st: left censoring

Hi

iam trying to estimate duration models(Cox-PH) but i have left
censoring..Does anyone knows how this is treated in STATA?

Ioanna Stylianou

----- Original Message -----
From: Christopher Baum <[hidden email]>
Date: Thursday, December 4, 2008 1:54 pm
Subject: st: Re: how to select model in xtpcse , corr(ar1)
To: [hidden email]


> < >
>  Ghislain said
>  I want to compare two models one made with : xtpcse x y z,corr(ar1)  
>
>  the other with : xtpcse x y w,corr(ar1) xtpcse doesn't return AIC nor

>  
>  BIC nor e(ll),
>  
>  Martin suggests looking at r^2.  AIC and BIC are criteria that will  
>
>  penalize for non-parsimonious models. As both of the models you  
>  specify have exactly the same length, they would not be useful. They

>
>  are not estimated with maximum likelihood, so there is no e(ll). [OLS

>  
>  is not estimated with MLE either, but -regress- does provide e(ll)].

>  
>  As r^2 is perfectly comparable for models of the same length, r^2  
>  would seem to make sense as a comparison.
>  
>  I would worry more about the notion that if each of these models does

>  
>  a good job, with a significant z or w respectively, then there is  
>  surely the risk that they are both misspecified versions of a model  
>
>  that contains both. Can you indeed rule out w appearing in the first

>
>  model, or z in the second?
>  
>  
>  Kit Baum, Boston College Economics and DIW Berlin
>  http://ideas.repec.org/e/pba1.html
>  An Introduction to Modern Econometrics Using Stata:
>  http://www.stata-press.com/books/imeus.html
>  
>  *
>  *   For searches and help try:
>  *   http://www.stata.com/help.cgi?search
>  *   http://www.stata.com/support/statalist/faq
>  *   http://www.ats.ucla.edu/stat/stata/
>  
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
Reply | Threaded
Open this post in threaded view
|

Re: st: RE: left censoring

IOANNA MARIOU STYLIANOU
No! that must be something new..i will though, thank you so much!


Ioanna

----- Original Message -----
From: "Mak, Timothy" <[hidden email]>
Date: Friday, December 5, 2008 6:18 am
Subject: st: RE: left censoring
To: [hidden email]


> Have you checked the -stset- help menu?
>  
>  Tim
>  
>  -----Original Message-----
>  From: [hidden email]
>  [mailto:[hidden email]] On Behalf Of IOANNA MARIOU
>  STYLIANOU
>  Sent: 04 December 2008 20:01
>  To: [hidden email]
>  Subject: st: left censoring
>  
>  Hi
>  
>  iam trying to estimate duration models(Cox-PH) but i have left
>  censoring..Does anyone knows how this is treated in STATA?
>  
>  Ioanna Stylianou
>  
>  ----- Original Message -----
>  From: Christopher Baum <[hidden email]>
>  Date: Thursday, December 4, 2008 1:54 pm
>  Subject: st: Re: how to select model in xtpcse , corr(ar1)
>  To: [hidden email]
>  
>  
>  > < >
>  >  Ghislain said
>  >  I want to compare two models one made with : xtpcse x y
> z,corr(ar1)  
>  >
>  >  the other with : xtpcse x y w,corr(ar1) xtpcse doesn't return AIC
> nor
>  
>  >  
>  >  BIC nor e(ll),
>  >  
>  >  Martin suggests looking at r^2.  AIC and BIC are criteria that
> will  
>  >
>  >  penalize for non-parsimonious models. As both of the models you  
>  >  specify have exactly the same length, they would not be useful. They
>  
>  >
>  >  are not estimated with maximum likelihood, so there is no e(ll). [OLS
>  
>  >  
>  >  is not estimated with MLE either, but -regress- does provide e(ll)].
>  
>  >  
>  >  As r^2 is perfectly comparable for models of the same length, r^2  
>
>  >  would seem to make sense as a comparison.
>  >  
>  >  I would worry more about the notion that if each of these models does
>  
>  >  
>  >  a good job, with a significant z or w respectively, then there is  
>
>  >  surely the risk that they are both misspecified versions of a
> model  
>  >
>  >  that contains both. Can you indeed rule out w appearing in the first
>  
>  >
>  >  model, or z in the second?
>  >  
>  >  
>  >  Kit Baum, Boston College Economics and DIW Berlin
>  >  http://ideas.repec.org/e/pba1.html
>  >  An Introduction to Modern Econometrics Using Stata:
>  >  http://www.stata-press.com/books/imeus.html
>  >  
>  >  *
>  >  *   For searches and help try:
>  >  *   http://www.stata.com/help.cgi?search
>  >  *   http://www.stata.com/support/statalist/faq
>  >  *   http://www.ats.ucla.edu/stat/stata/
>  >  
>  *
>  *   For searches and help try:
>  *   http://www.stata.com/help.cgi?search
>  *   http://www.stata.com/support/statalist/faq
>  *   http://www.ats.ucla.edu/stat/stata/
>  
>  *
>  *   For searches and help try:
>  *   http://www.stata.com/help.cgi?search
>  *   http://www.stata.com/support/statalist/faq
>  *   http://www.ats.ucla.edu/stat/stata/
>  
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
Reply | Threaded
Open this post in threaded view
|

Re: st: RE: left censoring

IOANNA MARIOU STYLIANOU
In reply to this post by Mak, Timothy
sorry, i mean there is a new option " intcens" which is something new..but iam not sure if it works..


ioanna

----- Original Message -----
From: "Mak, Timothy" <[hidden email]>
Date: Friday, December 5, 2008 6:18 am
Subject: st: RE: left censoring
To: [hidden email]


> Have you checked the -stset- help menu?
>  
>  Tim
>  
>  -----Original Message-----
>  From: [hidden email]
>  [mailto:[hidden email]] On Behalf Of IOANNA MARIOU
>  STYLIANOU
>  Sent: 04 December 2008 20:01
>  To: [hidden email]
>  Subject: st: left censoring
>  
>  Hi
>  
>  iam trying to estimate duration models(Cox-PH) but i have left
>  censoring..Does anyone knows how this is treated in STATA?
>  
>  Ioanna Stylianou
>  
>  ----- Original Message -----
>  From: Christopher Baum <[hidden email]>
>  Date: Thursday, December 4, 2008 1:54 pm
>  Subject: st: Re: how to select model in xtpcse , corr(ar1)
>  To: [hidden email]
>  
>  
>  > < >
>  >  Ghislain said
>  >  I want to compare two models one made with : xtpcse x y
> z,corr(ar1)  
>  >
>  >  the other with : xtpcse x y w,corr(ar1) xtpcse doesn't return AIC
> nor
>  
>  >  
>  >  BIC nor e(ll),
>  >  
>  >  Martin suggests looking at r^2.  AIC and BIC are criteria that
> will  
>  >
>  >  penalize for non-parsimonious models. As both of the models you  
>  >  specify have exactly the same length, they would not be useful. They
>  
>  >
>  >  are not estimated with maximum likelihood, so there is no e(ll). [OLS
>  
>  >  
>  >  is not estimated with MLE either, but -regress- does provide e(ll)].
>  
>  >  
>  >  As r^2 is perfectly comparable for models of the same length, r^2  
>
>  >  would seem to make sense as a comparison.
>  >  
>  >  I would worry more about the notion that if each of these models does
>  
>  >  
>  >  a good job, with a significant z or w respectively, then there is  
>
>  >  surely the risk that they are both misspecified versions of a
> model  
>  >
>  >  that contains both. Can you indeed rule out w appearing in the first
>  
>  >
>  >  model, or z in the second?
>  >  
>  >  
>  >  Kit Baum, Boston College Economics and DIW Berlin
>  >  http://ideas.repec.org/e/pba1.html
>  >  An Introduction to Modern Econometrics Using Stata:
>  >  http://www.stata-press.com/books/imeus.html
>  >  
>  >  *
>  >  *   For searches and help try:
>  >  *   http://www.stata.com/help.cgi?search
>  >  *   http://www.stata.com/support/statalist/faq
>  >  *   http://www.ats.ucla.edu/stat/stata/
>  >  
>  *
>  *   For searches and help try:
>  *   http://www.stata.com/help.cgi?search
>  *   http://www.stata.com/support/statalist/faq
>  *   http://www.ats.ucla.edu/stat/stata/
>  
>  *
>  *   For searches and help try:
>  *   http://www.stata.com/help.cgi?search
>  *   http://www.stata.com/support/statalist/faq
>  *   http://www.ats.ucla.edu/stat/stata/
>  
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/