Announcement

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

  • panel with heteroskedasticity autocorrelation and cross-dependence

    Hello everyone,
    I am planning to conduct a panel analysis with T=60 (time periods) and N=25 (cross-sectional units). After testing for them,
    I have addressed the issues of heteroskedasticity and autocorrelation by using robust standard errors (e.g., xtreg, fe robust or xtreg, fe vce(cluster country1)). However, I am wondering if there is a way to account for cross-sectional dependence as well. Can the xtscc command help me in this regard? I appreciate the assistance from anyone who can provide insights on this matter. Thank you.

  • #2
    First (=Primo, in Italian) item: if what follows applies to you, please note the local preference for real given and full names, which is well explained in the FAQ, that also provide instruction on how to fix your subscription. Thanks.
    That said, welcome to this forum.
    As you stated, you're dealing with a T>N panel dataset, -xtreg is not the way to go. In addition to -xtscc- , you may want to take a look at -xtgls- and -xtregar- too.
    Last edited by Carlo Lazzaro; 08 Jun 2023, 09:51.
    Kind regards,
    Carlo
    (Stata 18.0 SE)

    Comment


    • #3
      Dear Carlo,

      Thank you for your response. I greatly appreciate it. I would like to request your further assistance.

      I have implemented an xtscc regression, but I encountered an issue when attempting to conduct a Hausman test to choose between the random effects (RE) and fixed effects (FE) models. STATA displayed the following warning message: "Warning: chi2 < 0 ==> model fitted on these data fails to meet the asymptotic assumptions of the Hausman test; see suest for a generalized test."

      In an attempt to address this, I used the "suest" command with the "vce(cluster country)" option. However, STATA presented another error message stating, "Fixed was estimated with a nonstandard vce (Drisc/Kraay)."

      I kindly request your assistance in determining the appropriate test I should use to compare the RE and FE models.

      Thank you once again for your help.





      Comment


      • #4
        Davide:
        a tentative answer would point you out to the community-contributed module -rhausman-.
        Kind regards,
        Carlo
        (Stata 18.0 SE)

        Comment

        Working...
        X