```Date: Thu, 11 May 2006 03:05:20 -0400 Reply-To: Joe Whitehurst Sender: "SAS(r) Discussion" From: Joe Whitehurst Subject: Re: nonlinear panel data Comments: To: Dale McLerran In-Reply-To: <20060511001720.97436.qmail@web32203.mail.mud.yahoo.com> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Dale, I will address your mathematics a little later after we get rid of the annoying {which's) that I have indicated by enclosing your inappropriate use of( which 's ) with parentheses. Joe On 5/10/06, Dale McLerran wrote: > --- Jian Zhang wrote: > > > Hi: > > Is there a nonlinear counterpart for the "proc tscsreg" for panel > > data? Or does anyone have some idea about how to solve the > > heterogeneity and autocorrelation of the error term for nonlinear > > panel data? > > > > Thanks alot. > > > > Jian > > > > Jian, > > Can you assume a model of the form > > Y{ij} = f(X{ij},beta) + u{i} + eps{ij} > > u{i}~normal(0, V(u)) > cov(eps{i,j}, eps{i,k)) = rho**(abs(j-k)) > cov(eps{i,j}, eps{l,j}) = 0 > > That is, would you assume that, conditional on the fixed effects > (which) that enter nonlinearly and a random subject effect( which) that is > assumed normally distributed, then the residuals within a panel > follow an AR(1) covariance structure? > > If so, then you should be able to fit the model employing the > procedure NLMIXED. It won't be a simple, straightforward estimation > process. You would need all observations for a given panel on > a single row of your data. It would simplify writing of code if > your panels are all of equal length. However, that is not > necessary. > > Write back to the list with more details about your data and the > model that you would like to assume. > > Dale > > > --------------------------------------- > Dale McLerran > Fred Hutchinson Cancer Research Center > mailto: dmclerra@NO_SPAMfhcrc.org > Ph: (206) 667-2926 > Fax: (206) 667-5977 > --------------------------------------- > > __________________________________________________ > Do You Yahoo!? > Tired of spam? Yahoo! Mail has the best spam protection around > http://mail.yahoo.com > ```

Back to: Top of message | Previous page | Main SAS-L page