Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down at the end of May, and its replacement, statalist.org is already up and running.


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

Re: st: Fitting a linear regression with interval (inequality) constraints using nl


From   Nick Cox <njcoxstata@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Fitting a linear regression with interval (inequality) constraints using nl
Date   Tue, 14 Feb 2012 22:19:57 +0000

-nl- often benefits from careful initialisation of parameters.

Apart from that, the only easy thing I can say is that you have a
difficult problem.

Nick

On Tue, Feb 14, 2012 at 5:56 PM, Lieke Boonen (SiRM)
<Lieke.Boonen@sirm.nl> wrote:

> The problem we face is that we have to estimate a restricted OLS model for = which all betas have to be positive. We use the nl command (non lineair) in=  which we estimate the exp(beta) to make sure the beta is positive. However=  it takes a long time to converge and we want to know if there is an easier=  solution.
>
> In SAS you use the proc nlin command in which you state bounds for the beta=  to be larger than 0. You thus specify a normal lineair regression and add = the bound that the beta is larger than 0. In Stata we tried the constraint = regression (cnsreg) but we cannot create a constraint wich beta>0.
>
> Is there something we are missing or is the code we have written below the = only code that Works in such a situation?
>
> Thanks for your help.
>
> We have written the following code: (76 betas)
>
> nl (gemiddelde_kosten_gewogen =3D exp({lnb1})*lgnw1 + exp({lnb2})*lgnw2 + e=
> xp({lnb3})*lgnw3 + exp({lnb4})*lgnw4 + exp({lnb5})*lgnw5 ///
>            + exp({lnb6})*lgnw6 + exp({lnb7})*lgnw7 + exp({lnb8})*lgnw8 + e=
> xp({lnb9})*lgnw9 + exp({lnb10})*lgnw10 ///
>                                               + exp({lnb11})*lgnw11 + exp(=
> {lnb12})*lgnw12 + exp({lnb13})*lgnw13 + exp({lnb14})*lgnw14 + exp({lnb15})*=
> lgnw15 ///
>            + exp({lnb16})*lgnw16 + exp({lnb17})*lgnw17 + exp({lnb18})*lgnw=
> 18 + exp({lnb19})*lgnw19 + exp({lnb20})*lgnw20 ///
>                                               + exp({lnb21})*lgnw21 + exp(=
> {lnb22})*lgnw22 + exp({lnb23})*lgnw23 + exp({lnb24})*lgnw24 + exp({lnb25})*=
> lgnw25 ///
>            + exp({lnb26})*lgnw26 + exp({lnb27})*lgnw27 + exp({lnb28})*lgnw=
> 28 + exp({lnb29})*lgnw29 + exp({lnb30})*lgnw30 ///
>                                               + exp({lnb31})*ape1 + exp({l=
> nb32})*ape2 + exp({lnb33})*ape3 + exp({lnb34})*ape4 + exp({lnb35})*ape5 ///
>                                               + exp({lnb36})*ape6 + exp({l=
> nb37})*ape7 + exp({lnb38})*ape8 + exp({lnb39})*ape9 + exp({lnb40})*ape10 //= /
>                                               + exp({lnb41})*drempel_laag_= niet + exp({lnb42})*drempel_laag + exp({lnb43})*drempel_hoog_niet + exp({ln=
> b44})* drempel_hoog ///
>                                               + exp({lnb45})*mph + exp({ln= b46})*eph + exp({lnb47}) * avi11 + exp({lnb48})*avi21 + exp({lnb49})*avi31 =
> +exp({lnb50})* avi41 ///
>                                               + exp({lnb51}) * avi12 +exp(=
> {lnb52})*avi22 + exp({lnb53})*avi32 +exp({lnb54})* avi42 ///
>                                               + exp({lnb55}) * avi13 +exp(=
> {lnb56})*avi23 + exp({lnb57})*avi33 +exp({lnb58})* avi43 ///
>                                               + exp({lnb59}) * avi14 + exp=
> ({lnb60})*avi24 + exp({lnb61})*avi34 +exp({lnb62})* avi44 ///
>                                               + exp({lnb63}) * ses1a + exp= ({lnb64})*ses2a + exp({lnb65}) *ses3a + exp({lnb66})*ses4a + exp({lnb67}) *=  ses1b + exp({lnb68})*ses2b ///
>                                               + exp({lnb69}) *ses3b + exp(= {lnb70})*ses4b + exp({lnb71}) * fkg0 + exp({lnb72}) * fkg1 + exp({lnb73}) *=
>  fkg2 + exp({lnb74}) * fkg3 ///
>                                               + exp({lnb75}) * fkg4 + exp(=
> {lnb76}) * fkg5), noconst

*
*   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   |   Site index