Omitted variables in results using OLS model

[Patricia Ali]

Dear All,

I am analyzing data household level data and have included 10 provinces. I have also categorized my education variable into 1. household head no education 2. household head primary education 3. household head secondary education 4. household head higher education and 5. average education in the household

When i run my regressions, stata omits one province and also the no education variable and clearly states that this is due to collinearity. My VIF results though are below 10.

I assume that the omitted variables are not significant but would still like them displayed.

I am using the ols model with the command "reg Y X1 X2...."

How do i regress such that it still shows me all the variables or so that I get the standard error, coefficient and p value for the missing province and no education.

[FernandoRios]

Hi Patricia,
The short answer is NO.
If i understand your problem correctly, from your 10 provinces, 9 coefficients are estimated, and similarly, from your 3 education dummies, only 2 are displayed. There is no real work around it.
This is what is known as the dummy trap. You will always loose at least one dummy for each dummy set, because the full set of dummies are col-linear with the constant. If there is also collinearity within dummy sets (or other continuous explanatory variables), more dummies would be excluded.
The omitted categories are the base group, and one typically considers them as part of the constant.
HTH
Fernando

[Marcos Almeida]

You may also wish to check the "dependency among the independent variables" here.

[Patricia Ali]

Thanks Fernando and Marcos

[Carlo Lazzaro]

Patricia:
as an aside to previous excellent advice, you may want to take a look at https://en.wikipedia.org/wiki/Dummy_...le_(statistics).

[Patricia Ali]

Thanks Carlo