# [Limdep Nlogit List] Standarized tobit coefficients

William Greene wgreene at stern.nyu.edu
Mon Jan 4 13:16:10 EST 2010

```names; x=...,one\$ (Constant is last)
tobit; lhs=...;rhs=x\$
calc ; k=col(x)-1\$
matr ; vx=xvcm(x); vx=vx(1:k,1:k)\$
matr ; sx=diag(vx);sx=sqrt(vx)\$
matr ; beta=b(1:k)\$
calc ; sys=sqr(qfr(beta,vx)+s*s) \$
matr ; list ; std_beta=1/sys * sx * beta \$

/B. Greene

----- Original Message -----
From: "Macario Rodríguez Entrena" <macario.ent at gmail.com>
To: limdep at limdep.itls.usyd.edu.au
Sent: Sunday, January 3, 2010 10:39:11 AM GMT -05:00 US/Canada Eastern
Subject: [Limdep Nlogit List] Standarized tobit coefficients

Dear Limdep Users:

I am trying to obtain standardized tobit coefficients per Long (1997).
Initially, I calculated the standardized tobit coefficients after
estimation per Roncek (1992) as follows:

Bks(standarized)=Bk (Sigma k / Sigma y*)

Long (1997) criticized this approach because sigma is conditional on x.
He suggests instead to use the unconditional variance of y* computed
with the quadratic form:

(Sigma^ y*)2=B^' Var^(x) B + (Sigma^ E)2

where Var^(x) is the estimated covariance matrix among the x's and
(Sigma^ E)2 is the ML estimate of the variance of E.

Unfortunately, I am having difficulty with transferring this formula into
Limdep.
I hope someone on the list could offer some help.