Lab 04- Q2a

[malmufre]

Hello,

For Q2a, I am trying to run a regression table and I am getting the following:

n1.model <- lm(HealthStatus ~ 
PublicHousing +  
Race_1 +
Race_2 +
Race_3 +
Race_4 +
Education +
Age +
MaritalStatus_1 +
MaritalStatus_2 +
MaritalStatus_3 +
MaritalStatus_4 , data = data)

stargazer( n1.model,
           type = "text", 
           dep.var.labels = ("Health Status"),
           column.labels = "",
           covariate.labels = c("Public Housing","White","Black", "Hispanic", "Other" , "Education" , "Age", "Single", "Married", "Widow", "Divorced" ),
           digits = 2 )

which returns this:

The problem here is that I have 2 of my control variables(Other and Divorced) with no coefficients. Is it okay to have missing coefficients or should I be including only race and marital status without them being in detail? For example, I would include Race instead of Race_1, Race_2, Race_3, and Race_4?

Thank you!

[lecy]

The problem is that you can never include all levels of a factor in a regression model. One will always be omitted.

m <- lm( y ~ x + male_dummy )   # ok
m <- lm( y ~ x + female_dummy )   # ok
m <- lm( y ~ x + male_dummy + female_dummy )   # not ok

Since you did not select which one to omit, R is doing it for you.

It is better for you to select the omitted categories yourself. Remember that whichever group you omit will become the reference group represented by the intercept b0.

It's best to put the "typical" or reference group in the intercept as a point of comparison for the other variables.

Also recall that all of the coefficient tests reported in the table will be in relation to the omitted category. So if you want to compare married and divorced people in the model omit one of those groups. If you omit widowed then everything is compared to widowed and you won't be able to see if married and divorced groups differ from each other.