[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
st: bivariate probit
st: bivariate probit
Wed, 14 Apr 2010 22:56:58 -0700 (PDT)
I badly need help... Please help.. anybody who knows how to solve my dilemmas.
I am trying to run bivariate probit using STATA/SE 8.2. I would like verify when to use the bivariate probit and the seemingly unrelated bivariate probit regression which are the two bivariate options shown in the dropdown menu . What comes to my mind by checking the drop down options in STATA is to use bivariate probit if the two equations have the same set of independent variables while SUR Bivariate probit would be used when the independent variables for the two equations are not the same. My readings though somehow implied that I could still use bivariate probit even if the independent variables of the two equations are not exactly the same. The variables in the second equation which are not in the first equation are called the instruments (for selectivity problem). In this care, how can I run bivariate probit if the two equations didn’t have the same independent variables since in the bivariate probit window there’s only one box for the
Again, based on what I can remember from my readings, the SUR bivariate probit will be used if the dependent variable of the second equation is includes as one of the independent variables of the first equation. Is this right?
Another thing is, if I have 4 dummy variables for education levels (elementary, high school, some college, college graduate), how do I get marginals in such a way that elementary education will be used as a reference for the other levels, e.g., I would be able to say that having:
high school education is 3.1% less like to fall into low paid situation than elementary education
some college is 7.5% less like to fall into low paid situation than elementary education
college graduate is 15.2% less like to fall into low paid situation than elementary education
I was able to get the marginal/discrete change of each of these variables which I know, can be interpreted as the discrete probability as the value changes from zero to 1, e.g., I would be able to say:
1. Having elementary education is 15% less likely to fall into poverty than having no elementary education
2. Having high school education is 25.5% less likely to fall into poverty than having no high school education.
2. Having some college is 60.2% less likely to fall into poverty than having no college at all.
3. Being a college graduate has 80.5% less likelihood to fall in poverty than not a college graduate.
Again, what would I do if I want to compare elementary education with the rest of the categories?
One more thing is, what would be the layout of my data? Is it just okey to combine everything from year 1 to year 8 as long as I am not workng with panel data? I won't use the YEAR as a variable anyway.
We are still trying to come up with panel data. If we do have panel data, would the same layout be good still?
I highly appreciate anyone's help. Thanks so much.
* For searches and help try: