[Limdep Nlogit List] t-stat of Nested logit model using RU2 procedure

[William Greene]

Dear Chinh. Using the delta method, when you compute
a = 1/b, the estimator of the asymptotic variance for
a would be [da/db]^2 * Asy.Var[b] = [(-1/b^2)]^2 * Var[b].
Let s(b) = the square root of Var[b]. The t-ratio that you
compute for b is b/s(b).  The t ratio that you compute for
a will be (1/b) / [(1/b^2)*s(b)] = b/s(b). That is, the
t ratio is identical.
/Bill Greene

----- Original Message -----
From: "ho quoc Chinh" <hoqchinh at yahoo.com.vn>
To: "Limdep and Nlogit Mailing List" <limdep at limdep.itls.usyd.edu.au>
Sent: Wednesday, August 26, 2009 10:19:38 PM GMT -05:00 Columbia
Subject: [Limdep Nlogit List] t-stat of Nested logit model using RU2 procedure

Dear Limdep-users,
 
For pre-August 2003 versions of Nlogit, there is a problem of IV, namely the reported IV value which is actually the scale parameter of elemental alternative level, i.e.,mu. It is easy to obtain the real IV of two-level NL models by taking the reciprocal of the reported IV value. However, how to calculate the t-statatistic for real IV value? Is it possible to use reported t-stat of corresponding RU1 procedure? My NL model sets all attribute parameters to be alternative-specific between partitions.
 
Any suggestion is highly appreciated.

Best regards,
Chinh.


      Kết nối để chat trên blog hoặc bất kỳ trang web cá nhân nào! Yahoo! cho phép bạn có thể chat ngay với Pingbox. Thử xem! http://vn.messenger.yahoo.com/pingbox/
_______________________________________________
Limdep site list
Limdep at limdep.itls.usyd.edu.au
http://limdep.itls.usyd.edu.au