Announcement

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

  • Few clusters

    Hi,

    I am trying to estimate a model with clustered errors, however the number of clusters is quite low, approximately 10.

    I have checked several references about it. I was wondering whether adding an option vce(cluster variable) is sufficient, given that STATA takes into account the T distribution to derive the p values.

    Could you suggest how I can take into account and correct for the low number of clusters?

    Thank you in advance.

    Best,

    Nikos

  • #2
    Nikos:
    you should provide some more details about the reason supporting your clustering.
    Theoretically speaking, 10 clusters are too few to justify a sensible clustering.
    Kind regards,
    Carlo
    (Stata 18.0 SE)

    Comment


    • #3
      Thanks for your answer. I am conducting a analysis using survey data from several countries, and I have found some references which essentially correct for small number of clusters. However, I am quite confused with the appropriate Stata coding.

      http://jhr.uwpress.org/content/50/2/317.short

      https://www.mitpressjournals.org/doi.../rest.90.3.414

      Comment

      Working...
      X