Announcement

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

  • Panel Data Estimator Confusion

    Hi Statalist,

    I'm currently attempting to look at the impact of different variables on democracy. I am using the V-dem score for democracy which is a continuous variable between 0-1 (v2x_libdem). For independent variables I am using GDP per capita, GDP growth, life expectancy, primary school enrollment and cellular coverage. I have been reading a lot online to find the best model for undertaking this task. I have 42 different countries and a range between 57 and 4 years for each country. This gives me 1,987 observations.

    Models I have looked at include:

    xtreg v2x_libdem gdp gdpgpc lifeexp prisch mobile i.year, fe

    xtregv2x_libdem gdp gdpgpc lifeexp prisch mobile, vce(robust)

    xtabond v2x_libdem gdp gdpgpc lifeexp prisch mobile, vce(robust)

    xtgls v2x_libdem gdp gdpgpc lifeexp prisch mobile

    and looking into xtpcse.

    However, I just can't work out which one is best for unbiased and most efficient coefficients.

    Best,

    Charlie


  • #2
    Charlie:
    as you have a long N, long T panel dataset, you may want to consider, alternatively: -xtgls-; -xtregar, fe- or -xtregar, re-.
    As an aside, at its face value, your query proposed many different panel data regression models and some of them are really different.
    An unsolicited advice would highlight the need of a deep awarenes/knowledge of the different models (that implies taking more than one look at any decent panel data econometrics textbook) you listed in your post, being aware that -click and see- is by far the most dangerous approach in statistics.
    Kind regards,
    Carlo
    (Stata 19.0)

    Comment

    Working...
    X