Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Check if xtnbreg, fe is robust

    Hi,

    I am working with STATA/SE 17.0. I have a panel data for 10 countries for the 2000-2021 period.

    My dependent variable is a count variable, and the variance is much bigger than the mean.

    I decided to use the negative binomial model.

    I have a few questions:

    1) Do I need to check for stationarity before doing the regression? If not can you tell me any literature where I can read about stationarity on count data?

    2) What do you advise to do to check if the results are robust?

    The code I have is the follow.

    ​​​​​​// Import your data
    import excel using "TOP10_efficiency.xlsx", firstrow clear
    save "Data_efficiency.dta", replace // save data

    // Declare it as panel data
    xtset ID year, yearly // declare it to be panel data
    duplicates report ID year

    // Descriptive Statistics
    summarize

    // Dependent variable details
    sum patentsEFF, detail

    // Correlation matrix
    correlate patentsEFF patentsRE EPSindex oilprice RE_RD fossilfuelsrent gdpcapita PARIS

    // Creating the multiplicative dummy for the years 2015-2021
    generate year_dummy = (year >= 2015 & year <= 2021)
    generate gdp_per_capita_x_year_dummy = gdpcapita * year_dummy

    // Principal regression
    xtnbreg L.patentsEFF L.patentsRE EPSindex oilprice RE_RD fossilfuelsrent gdp_per_capita_x_year_dummy gdpcapita PARIS, fe noconstant

    Thank you

  • #2
    See #3 on why you should almost never use the Negative Binomial FE estimator: https://www.statalist.org/forums/for...-poisson-model

    Comment

    Working...
    X