Hello, I'm new to this forum.
I'm struggling with my stata code.
I'm using is a collection of five years of data, where dependent variable is mortality, independent variable is number of doctor in a county, and the unit is county (there are 229 counties in my data) - pooled cross-sectional data (It's not panel data because I collected data from various sources)
I wanna do time series poisson regession but I'm not sure if my code is right.
My code is:
poisson mort doctor_density bed female eld i.year, vce(robust) irr
The model is
: ln(mortit)= α+ β(doctor_densityit)+ γCovariateit + Tt
It's the simplest code that I thought.
I heard that there's a way not to use year as a dummy variable, but don't don't know how...
# If there's any better way to deal with time, please let me know.
Another code I thought is:
tsset county year, yearly
poisson mort l.mort doctor_density bed female eld, vce(robust) irr
The model is
: ln(mortit)= α+ β1ln(morti(t-1))+β2(doctor_densityit)+ γCovariateit
I used the method used for panel data and found that cross-sectional data can be analyzed by considering it as pseudo-panel data.
If I consider it as pseudo-panel data, it's perfectly balanced so that I think the idea is quite good.
I wanted to used lagged option that can be used for panal data. (mortality is generally associated with last year's mortality)
And I'm wondering - # if I use lagged option with poisson regression, the mortality of (t-1) is logged as well.
# Is there any other consideration for using pseudo-panel anlaysis?
Thank you in advance
Hans
I'm struggling with my stata code.
I'm using is a collection of five years of data, where dependent variable is mortality, independent variable is number of doctor in a county, and the unit is county (there are 229 counties in my data) - pooled cross-sectional data (It's not panel data because I collected data from various sources)
I wanna do time series poisson regession but I'm not sure if my code is right.
My code is:
poisson mort doctor_density bed female eld i.year, vce(robust) irr
The model is
: ln(mortit)= α+ β(doctor_densityit)+ γCovariateit + Tt
It's the simplest code that I thought.
I heard that there's a way not to use year as a dummy variable, but don't don't know how...
# If there's any better way to deal with time, please let me know.
Another code I thought is:
tsset county year, yearly
poisson mort l.mort doctor_density bed female eld, vce(robust) irr
The model is
: ln(mortit)= α+ β1ln(morti(t-1))+β2(doctor_densityit)+ γCovariateit
I used the method used for panel data and found that cross-sectional data can be analyzed by considering it as pseudo-panel data.
If I consider it as pseudo-panel data, it's perfectly balanced so that I think the idea is quite good.
I wanted to used lagged option that can be used for panal data. (mortality is generally associated with last year's mortality)
And I'm wondering - # if I use lagged option with poisson regression, the mortality of (t-1) is logged as well.
# Is there any other consideration for using pseudo-panel anlaysis?
Thank you in advance
Hans
