Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

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

Re: st: Conditional Probabilities after bivariate probit

From   Austin Nichols <>
Subject   Re: st: Conditional Probabilities after bivariate probit
Date   Fri, 10 Jun 2011 13:30:05 -0400

orc un <>:
Don't use -mfx-
just read
to make any probability you like:

webuse school, clear
biprobit (private = logptax loginc years) (vote = logptax years)
predict double p11   , p11
predict double p10   , p10
predict double p01   , p01
predict double p00   , p00
predict double pmarg1, pmarg1
predict double pmarg2, pmarg2
predict double pcond1, pcond1
predict double pcond2, pcond2
g double p_1=(p11+p01)/pmarg2
g double p1_=(p11+p10)/pmarg1
su p_1 p1_
g double cp1=p11/(p11+p01)
g double cp2=p11/(p11+p10)
su cp1 pcond1
su cp2 pcond2
g double zp1=p10/(p10+p00)
la var zp1 "Pr(y1==1|y2==0)"
g double zp2=p01/(p01+p00)
la var zp1 "Pr(y2==1|y1==0)"

On Fri, Jun 10, 2011 at 12:48 PM, orc un <> wrote:
> Dear Statalist users,
> I have a question regarding the conditional probability computation for Bivariate Probit model. As far as I know it is possible to compute P(Y1=1|Y2=1) for a bivariate Probit model using the standard Stata commands as mfx, predict(pcond1) .However I would like to compute the P(Y1=1|Y2=0). Is it possible to compute this marginal effects by just creating an alternative dependent variable(Y2c) which takes the complement values of Y2 (ie for Y2=1 then Y2c=0 and for Y2=0 then Y2c=1) and compute P(Y1=1|Y2c=1) instead of P(Y1=1|Y2=0). If yes is it possible to use the same approach for recursive bivariate probit  model??

*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index