Announcement

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

  • Issue with three-dimensional panel data analysis

    Hi,

    I am new with STATA and I would like some advice for the following problem: I am dealing with panel count data model.My dependent variable is a count and is the number(count) of investment projects in each host country(i), in each sector (j) in a given year(t). My data is panel data from 2003 to 2016 and I have 12 industries and 105 host nations. My two main explanatory variables vary by industry by time(ij)and by country, industry, time(ijt).Control variables are composed of i and t. Therefore,
    I am dealing with three-dimensional panel data analysis:
    i= country, 105
    j= industry, 12
    t =year, 14
    I am using xtpoisson with fe approach with robust standard errors.

    After reading a lot of STATAlist previous posts I realized that in order to xtset my data I need to combine countries*industries fixed effect:
    egen panelid= group (country * industries)
    xtset panelid Year
    xtset Y X i.Year, robust i(panelid) fe

    However, because I have a lot of countries (105) and a lot of zeros on my dependent variable an important downfall of this estimation is the loss of degrees of freedom because of the inclusion of all these dummy variables. Instead of interacting countries* industries and because I don’t want to combine the country and industry FE, I also tried to put them separately on the model:

    For a model with industry FE:
    Xtset Industry
    Xtpoisson Y X i. Year i.country, fe
    Xtpoisson Y X. i.Year i.region, fe
    I also incorporate regional dummies in order to group my 105 countries and I incorporated in the model:

    Xtpoisson Y X. i.Year i.region, fe


    My question is that is there any other way to model a three dimensional panel data without combining industries and countries which generates so many dummies? At the same time when I am grouping industries & countries does not allow me to have a separate information about industries or countries.

    I found this older post that was helpful on my decision.

    https://www.statalist.org/forums/for...ata-regression

    Any suggestions or advice will be greatly appreciated.

    Thank you very much,



  • #2
    I would probably do this as:
    Code:
    xtset country
    xtpoisson Y X i.industry i.year, fe
    There does not appear to be any need for you to specify a time variable in your -xtset- command, because you are not using time series operators (such as lag, lead, etc.), nor are you fitting models with autoregressive correlation structure. So skip the time variable in -xtset- and then you don't have to worry about unique observations.

    I don't know what to say about region, which you only mention late in your description. I can't even tell if regions are within countries or are groups of countries. Either way, there is nesting there and in a fixed-effects framework you will not be able to simultaneously estimate the effects at both levels (though adjusting for whichever is at the lower level of nesting will automatically adjust for the other). If estimating separate country and region effects are important to your research goals you will have to abandon fixed-effects modeling.

    Comment


    • #3
      Dear Clyde Schechter,

      Thank you very much for your reply, which is very helpful.
      I wanted to check if by skipping the time serious structure on my panel is not wrong. Thanks for your advice on that.

      I though about doing that as well. When I tried that all my control variables that vary by country by time become statistical insignificant.I pressume this is happening because that involves a considerable loss of degrees of freedom since I have 105 countries and many zeros(91% on my dependent variable which is count).
      With regards to regional dummies, I incorporated regions in a way to group my countries based on previous literature. Definitely estimating seperately country and region effect is not important for me.I just thought to incorporate regions (as more aggregate groups of my countries) in a way to avoid including invividual country specific dummies.
      My main research focus is the industries and this is why I used-xtset industry-.However, industries are nested within countries.


      Thank you for your time.

      Comment

      Working...
      X