Statalist


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

[no subject]



The critical values for the 2SLS relative bias with two endogenous regressors are only available if you have at least four excluded instruments. With one endogenous regressor, the critical values are only available if you have at least three excluded instruments. 

--- On Tue, 4/14/09, Jennifer Beardsley <jenn_beardsley@yahoo.com> wrote:

> From: Jennifer Beardsley <jenn_beardsley@yahoo.com>
> Subject: Re: st: Re: estat firststage blank results
> To: "Stata  list" <statalist@hsphsun2.harvard.edu>
> Date: Tuesday, April 14, 2009, 2:52 PM
> Dear Martin, I don't know why my
> message below didn't post, so I'll try again (see further
> below).
> 
> Dear Brian, many thanks for your response as well. However,
> you said that the 2sls relative bias statistic is available
> for 1 or 2 endogenous regressors, not for any more. Note
> though that in one of my main regressions, I have only 2
> end. regressors, but no 2sls relative bias statistic. My
> equation looks like this: 
> 
> . ivregress 2sls
> > lnexpendpa sex_x age_ age2 adult_equiv dist_healthcent
> primary_enroll 
> > (road_per_t_pop yield_temp = rain rain_cv )
> > if  year==2000
> 
> And the -estat firststage- results look like this:
> 
> 
> . estat firststage
> 
>   Shea's partial R-squared
>   --------------------------------------------------
>            
>    |     Shea's 
>            Shea's
>       Variable |  Partial
> R-sq.   Adj. Partial R-sq.
>   -------------+------------------------------------
>   road_per_t~p |     0.0065 
>            0.0061
>     yield_temp | 
>    0.0057         
>    0.0053
>   --------------------------------------------------
> 
> 
>   Minimum eigenvalue statistic = 31.4186 
>    
> 
>   Critical Values         
>             # of endogenous
> regressors:    2
>   Ho: Instruments are weak       
>      # of excluded instruments: 
>    2
>  
> ---------------------------------------------------------------------
>                
>                
>      |    5% 
>    10%     20% 
>    30%
>   2SLS relative bias         
>        |     
>    (not available)
>  
> -----------------------------------+---------------------------------
>                
>                
>      |   10% 
>    15%     20% 
>    25%
>   2SLS Size of nominal 5% Wald test 
> |   7.03    4.58   
> 3.95    3.63
>   LIML Size of nominal 5% Wald test 
> |   7.03    4.58   
> 3.95    3.63
>  
> ---------------------------------------------------------------------
> 
> Any idea why even 2 endog regressors won't produce a 2sls
> relatie bias statistic?
> 
> Best,Jennifer
> 
> ------------------------------------------
> From   "Brian P. Poi" <bpoi@stata.com>
> 
> To   statalist@hsphsun2.harvard.edu
> 
> Subject   Re: st: estat firststage blank
> results 
> Date   Tue, 14 Apr 2009 08:20:39 -0500 (CDT)
> 
> 
> Jennifer,
> 
> The critical values for the minimum eigenvalue statistic
> are based on tables of results published in Stock and Yogo
> (2005). For the 2SLS and LIML sizes of nominal 5% Wald
> tests, Stock and Yogo only considered the cases where the
> model has one or two endogenous regressors; since you have 3
> endogenous regressors, critical values are not available.
> For the 2SLS relative bias with 3 endogenous regressors,
> Stock and Yogo's table only contains critical values for
> models with at least 5 excluded instruments. 
> 
>    -- Brian Poi
>    -- bpoi@stata.com
> 
> --- On Mon, 4/13/09, Jennifer Beardsley <jenn_beardsley@yahoo.com>
> wrote:
> 
> 
> From: Jennifer Beardsley <jenn_beardsley@yahoo.com>
> Subject: st: Re: estat firststage blank results
> To: "Stata list" <statalist@hsphsun2.harvard.edu>
> Date: Monday, April 13, 2009, 9:31 PM
> 
> Dear Martin,
> Interesting, I had not known about this -trace- command. I
> followed your suggestion, below is the area where "not
> available" shows up (it shows up three times). I bolded and
> redded it, but not sure if formatting is retained on
> statalist. If not, a simple ctrl-search can find the 3
> instances of "not available". Looking at the output, it
> looks quite complex to me, please let me know if you can
> interpret from this.
> Thanks,
> Jennifer
>  
> Minimum eigenvalue statistic = 8.50405 
> - di
> 
> - di as text _col(3) "Critical Values" _col(40) "# of
> endogenous regressors:" as 
> > res %5.0f `nendog'
> = di as text _col(3) "Critical Values" _col(40) "# of
> endogenous regressors:" as 
> > res %5.0f 3
> Critical Values # of endogenous regressors: 3
> - di as text _col(3) "Ho: Instruments are weak" _col(40) "#
> of excluded instrumen
> > ts: " as res %5.0f `nexclexog'
> = di as text _col(3) "Ho: Instruments are weak" _col(40) "#
> of excluded instrumen
> > ts: " as res %5.0f 4
> Ho: Instruments are weak # of excluded instruments: 4
> - di as text _col(3) "{hline 35}{c TT}{hline 33}"
> ---------------------------------------------------------------------
> - di as text _col(38) "{c |}" _col(43) "5% 10% 20% 30%"
> | 5% 10% 20% 30%
> - di as text _col(3) "2SLS relative bias" _col(38) "{c |}"
> _c
> 2SLS relative bias | - if `mineigcv'[1,1] < . {
> = if __00000H[1,1] < . {
> di as res _col(41) %5.2f `mineigcv'[1,1] _col(49) %5.2f
> `mineigcv'[1,2] _col(57
> > ) %5.2f `mineigcv'[1,3] _col(65) %5.2f
> `mineigcv'[1,4]
> }
> - else {
> - di as text _col(48) "(not available)"
> (not available)
> - }
> - di as text _col(3) "{hline 35}{c +}{hline 33}"
> -----------------------------------+---------------------------------
> - di as text _col(38) "{c |}" _col(42) "10% 15% 20% 25%"
> | 10% 15% 20% 25%
> - di as text _col(3) "2SLS Size of nominal 5% Wald test"
> _col(38) "{c |}" _c
> 2SLS Size of nominal 5% Wald test | - if `mineigcv'[2,1]
> < . {
> = if __00000H[2,1] < . {
> di as res _col(41) %5.2f `mineigcv'[2,1] _col(49) %5.2f
> `mineigcv'[2,2] _col(57
> > ) %5.2f `mineigcv'[2,3] _col(65) %5.2f
> `mineigcv'[2,4]
> }
> - else {
> - di as text _col(48) "(not available)"
> (not available)
> - }
> - di as text _col(3) "LIML Size of nominal 5% Wald test"
> _col(38) "{c |}" _c
> LIML Size of nominal 5% Wald test | - if `mineigcv'[2,1]
> < . {
> = if __00000H[2,1] < . {
> di as res _col(41) %5.2f `mineigcv'[3,1] _col(49) %5.2f
> `mineigcv'[3,2] _col(57
> > ) %5.2f `mineigcv'[3,3] _col(65) %5.2f
> `mineigcv'[3,4]
> }
> - else {
> - di as text _col(48) "(not available)"
> (not available)
> - }
> - di as text _col(3) "{hline 35}{c BT}{hline 33}"
> ---------------------------------------------------------------------
> 
>  
> 
> 
> 
> 
> From
>   "Martin Weiss" <martin.weiss1@gmx.de>
> 
> To
>   <statalist@hsphsun2.harvard.edu>
> 
> Subject
>   st: Re: estat firststage blank results
> 
> Date
>   Tue, 14 Apr 2009 02:25:47 +0200
> 
> <>
> 
> 
> -set trace on- with -set tracehilite "not available"- and
> report the area where the phrase appears in bold... HTH
> Martin
> _______________________
> 
> ----- Original Message ----- From: "Jennifer Beardsley"
> <jenn_beardsley@yahoo.com>
> To: <statalist@hsphsun2.harvard.edu>
> Sent: Tuesday, April 14, 2009 2:18 AM
> Subject: st: estat firststage blank results
> 
> 
> 
> Dear Statalist,
> 
> When conducting an IV regression using -ivregress 2sls- and
> following that up to get statistics indicating the strength
> of the instruments, using -estat firststage-, the second
> half of the latter's results don't provide any statistics.
> In particular, it looks as follows:  Minimum eigenvalue
> statistic = 8.91958
>  Critical Values           
>           # of endogenous
> regressors:    3
>  Ho: Instruments are weak         
>    # of excluded instruments: 
>    3
> 
> ---------------------------------------------------------------------
>                
>                
>     |    5% 
>    10%     20% 
>    30%
>  2SLS relative bias         
>        |     
>    (not available)
> 
> -----------------------------------+---------------------------------
>                
>                
>     |   10% 
>    15%     20% 
>    25%
>  2SLS Size of nominal 5% Wald test  |   
>      (not available)
>  LIML Size of nominal 5% Wald test  |   
>      (not available)
> 
> Why may that be? There is no clue to this from the stata
> manual.
> Thanks for any thoughts,
> Jennifer
> 
> 
> 
> 
> 
> 
> 


      


*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index