[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: Simulatenous equation model with xtprobit

From   "C.T.Shehzad" <[email protected]>
To   [email protected]
Subject   Re: st: Simulatenous equation model with xtprobit
Date   Fri, 22 Aug 2008 17:21:39 +0200

Dear Stas:

Yes, the difference between categories is same and y1 and y2 (actual values ) should cause x but they can be endogenous and that's why I am using simulateneous equation model. Unfortunately, I donot have access to the book mentioned by Maatren for one or two weeks and I have to rely solely on gallam manual for the time being.

As an alternate approach i can also use ivprobit but that doesnot account for panel structure of data.

Best regards,


Stas Kolenikov wrote:

so if y's are ordinal, is the distance between categories 1 and 2 of
y1 the same as the distance between categories 2 and 3? That's what
your model explicitly assumes. Does your x depend on the realized
values of y*, or on their continuous propensity counterparts?

-gllamm- will do pretty much anything, you need to rearrange your data
in a marginally tricky way, and form appropriate -link-s.

On Fri, Aug 22, 2008 at 5:49 AM, Choudry T Shehzad <[email protected]> wrote:

Thanks Maarten... gllamm seems relevant but I have a question again. My
model is like

x = a + b1*y1 + b2*y2 + b3*y3
y1 = a2 + b4*y4 + b5* y3
y2 = a3 + b6*y6 + b4*y4

where x is binary and y1 and y2 are ordered. This is balanced panel
dataset (countries and years) and I want to run simultaneous equation
model becuase I suspect y1 and y2 are endogenous. Are you suggesting me
to use structural equation modelling via gllamm? If the data had been
continuous, I would have gone for reg3 but now after reading gllamm
manual, I still could not get a clear idea that how can I model y1 and y2.

*   For searches and help try:

© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index