RE: st: OLS versus Ordered Probit

[Cameron McIntosh <cnm100@hotmail.com>]

Vinod, 

Well, usually analyzing coarse scales with OLS (or any method that assumes a continuously distributed observed response) will give you attenuated estimates (but the difference tends to vanish as the number of response categories increase). See:

Peel, M.J., Goode, M.M.H, & Moutinho, L.A. (1998). Estimating consumer satisfaction: OLS versus ordered probability models. International Journal of Commerce and Management, 8(2), 75-93.

Beauducel, A., & Herzberg, P.Y. (2006). On the Performance of Maximum Likelihood Versus Means and Variance Adjusted Weighted Least Squares Estimation in CFA. Structural Equation Modeling, 13(2), 186-203. 

However, such methods can be computationally intensive (requiring several dimensions of integration with complex models), especially in the case when latent variables are involved. You may want to see a synthesis of probit and OLS methods (POLS):

van Praag, B.M.S., & Ferrer-i-Carbonell, A. (2006). An Almost Integration-free Approach to Ordered Response Models. Tinbergen Institute Discussion Paper, TI 2006-047/3. http://www.tinbergen.nl/discussionpapers/06047.pdf

Cam

> From: Vinod.Venkiteshwaran@tamucc.edu
> To: statalist@hsphsun2.harvard.edu
> Subject: st: OLS versus Ordered Probit
> Date: Fri, 9 Dec 2011 02:59:04 +0000
> 
> This question is not about stata per se but more of an econometrics query. 
> What are the strengths/drawbacks of using OLS, as opposed to ordered probit or logit, to estimate a model of ordered choices?
> 
> Thank you
> 
> Vinod
> 
> 
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
 		 	   		  
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/