Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

[no subject]

>Suppose you have a data set where x1, ..,xp are your
independent variables
>(including dummies for treatment(s)). Then for logistic
>coefficients c0, c1, c2,..,cp  issue the following
>gen z = c0 + c1*x1 +.... + cp*xp
>gen p = exp(x)/(1+exp(z))
>gen y=uniform()<=p
>z is the "xb", the linear predictor
>p is the probability that y=1
>The last statement generates random values of y that are 0
with probability
>1-p and 1 with probability p.
>It's up to you to set up the data set with the x's and to
specify the values
>of the coefficients (the c's). They can be eiher Stata
variables (a waste of
>space) or scalars or local macros.

1. What is the reason behind the use of the linear predictor
(-xb-) to obtain 'index' values?
2. Why the -uniform()- to generate random numbers? I know
it's a latent LHS variable, but don't we want
'deterministic' results?
3. What 'Stata variables' do you mean as vlaues for

Two days ago, Nick Cox suggested a similar approach
for -mlogit- (comments added):

qui forval i = 1/`e(k_cat)' {
	predict ypb_`i', outcome(`i') xb /* NOTE -xb- ! */
	replace ypb_`i' = exp(ypb_`i')

Are the Al's and Nick's approach identical (apart from
different amount of outcomes)?

For any clarifications I would be grateful,

-----Original Message-----
From: Sherry Weitzen []
Sent: Thursday, November 21, 2002 3:56 PM
Subject: st: simulating data

I feel very naive.. I'd like to simulate data from a
logistic model with
several independent variables (including a treatment
variable and several
confounders).  I have no idea where to start.  Please
advise.  thank you.

Sherry Weitzen
Department of Community Health
Brown University, Box G
Providence RI 02912
Reply to:

*   For searches and help try:
*   For searches and help try:

*   For searches and help try:

© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index