Re: st: linear probability model (LPM)

[Steven Samuels <ssamuels@albany.edu>]

A linear probability model is desirable because  effects are risk  
differences, which are much easier to interpret than odds ratios.  
It's best  for proportions that are not too close to 0 or 1;  
otherwise the model may predict probabilities outside those  
boundaries.  (In this range linear, probit, and logit models give  
similar predictions-Cox, Analysis of Binary Data, 1972).

In Stata you can use the -reg- command with the "robust" option to  
produce proper standard errors.  However you still need to check  
goodness of fit. With continuous covariates, use the -linktest-  
command.  You should also group  predictors, continuous or  
categorical,  to visually see if the linear model is accurate. With  
continuous covariates, the -lowess- command, with the -noweight- and - 
adjust- options will give you a visual check of fit without grouping.

Good luck!

Steve

On Apr 26, 2007, at 10:12 AM, Vanessa Mahlberg wrote:


> My dependent variable is dichotomous (zero for low work  
> satisfaction and one for high work satisfaction). I would like to  
> run a linear probability model (LPM) instead of a probit model- but  
> I don?t really know the right stata command. I read in some  
> articles that I should use the "regress" command?!? But in my  
> opinion it must be something like STATISTICS- BINARY OUTCOMES-???
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/