I have pooled cross-sectional survey data of NSSO rounds. The dependent variable has a binary outcome with 6 categorical independent variables. Now, Can you please suggest the correct way of doing regression analysis using stata?

Hi Sadaf,
For this kind of data, if your categorical variables are ordinal you need to perform ordinal regression while for nominal convert the data into dummy variable form by assigning values like 1 for manager, 2 – team lead and so on. Then perform logistic regression with dummy variables.

How would I deal with a scenario where some of my explanatory variables are not stationary as above? I understand taking the first difference is most suitable but when I attempt to do so in Stata, I receive “not ordered”. I take this as Stata thinking I am confused and implying that I should difference my variables prior to using xtset? Any help would be appreciated

Hey Ciaran
Command xtset needs to be used prior to any other calculation. The error “not ordered” might be due to misspecification of the panel data. Try to re-set the series as panel data and once check the type of variable is clearly specified. Run the stationarity test at the level and then at first difference.

2 years & 11 months ago

Hi,

I have pooled cross-sectional survey data of NSSO rounds. The dependent variable has a binary outcome with 6 categorical independent variables. Now, Can you please suggest the correct way of doing regression analysis using stata?

^{}2 years & 11 months ago

Hi Sadaf,

For this kind of data, if your categorical variables are ordinal you need to perform ordinal regression while for nominal convert the data into dummy variable form by assigning values like 1 for manager, 2 – team lead and so on. Then perform logistic regression with dummy variables.

4 years & 5 months ago

How would I deal with a scenario where some of my explanatory variables are not stationary as above? I understand taking the first difference is most suitable but when I attempt to do so in Stata, I receive “not ordered”. I take this as Stata thinking I am confused and implying that I should difference my variables prior to using xtset? Any help would be appreciated

^{}4 years & 5 months ago

Hey Ciaran

Command xtset needs to be used prior to any other calculation. The error “not ordered” might be due to misspecification of the panel data. Try to re-set the series as panel data and once check the type of variable is clearly specified. Run the stationarity test at the level and then at first difference.