Ordered Probit, Fitted values or Residuals?
Posted: Fri Jun 04, 2010 5:00 pm
Hello:
My issue involves estimating a model where the dependent variable (ALLCBI) is graphed in the attached word doc. This variable starts with a signed binary variable, -1,0,+1, but adds up these values if more than one occurs on one day, resulting in the data below, which ranges from -2 to +3. I am thinking this is an ordered probit (but could be wrong here), however, I need to use the residuals (or forecasts to create residuals) of this equation in a 2-step procedure in another model. The problem then is I see no way to get ordinary residuals from this model, and forecasts are not allowed. Following what the program indicates “Use Make Model”, the box indicates an error because the variable ALLCBI is used as and endogenous variable in the probit model? Using a simple linear probability model, ie, OLS, I get pretty good results both for fit and residuals, but this mode is probably not the best. Any information would be appreciated. I am using version 6.
Thank you.
Sbarnhar
My issue involves estimating a model where the dependent variable (ALLCBI) is graphed in the attached word doc. This variable starts with a signed binary variable, -1,0,+1, but adds up these values if more than one occurs on one day, resulting in the data below, which ranges from -2 to +3. I am thinking this is an ordered probit (but could be wrong here), however, I need to use the residuals (or forecasts to create residuals) of this equation in a 2-step procedure in another model. The problem then is I see no way to get ordinary residuals from this model, and forecasts are not allowed. Following what the program indicates “Use Make Model”, the box indicates an error because the variable ALLCBI is used as and endogenous variable in the probit model? Using a simple linear probability model, ie, OLS, I get pretty good results both for fit and residuals, but this mode is probably not the best. Any information would be appreciated. I am using version 6.
Thank you.
Sbarnhar