Announcement

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

  • xtivreg (first stage) and estout (vs. xtivreg2)

    Is it possible to generate latex code for the first stage of xtivreg (Stata 14) using estout (or otherwise)?

    By reading other related Q&A, it seems that xtivreg2 is the only option to programmatically produce latex for first stage estimation (rather than manually creating latex after each code run). Am I missing something?


    The PROBLEM with xtivreg2 is:

    Code:
    xtivreg , fe vce(robust)
    produces DIFFERENT standard errors, p-values, and confidence interval compared to:

    Code:
    xtivreg2 , fe robust
    PS: I would appreciate it if Mark Schaffer or someone from Stata could shed light on the differences in their VCE estimators and the underlying assumptions.

    PSS: The following discussion between Ariel Karlinsky and Emad Shehata seems relevant: http://www.statalist.org/forums/foru...ivreg-xtivreg2 ... Emad Shehata suggestion to use "small" still does not estimate the same SE (a point also noted by Emad Shehata) in the same thread ... finally, Ariel Karlinsky seems to imply without further references that xtivreg2 may have more accurate SE estimates

  • #2
    UPDATE: it seems that using "cluster' VCE for xtivreg2 results in very similar estimates. So, the SE should be compared for:

    IDENTICAL xtivreg:
    Code:
    xtivreg , fe vce (robust)
    xtivreg, fe vce(cluster clustervar)
    COMPARE TO THE FOLLOWING xtivreg2:
    Code:
    xtivreg2 , fe cluster(clustervar)




    Comment


    • #3
      Originally posted by Baykaar Kee View Post
      Is it possible to generate latex code for the first stage of xtivreg (Stata 14) using estout (or otherwise)?

      By reading other related Q&A, it seems that xtivreg2 is the only option to programmatically produce latex for first stage estimation (rather than manually creating latex after each code run). Am I missing something?


      The PROBLEM with xtivreg2 is:

      Code:
      xtivreg , fe vce(robust)
      produces DIFFERENT standard errors, p-values, and confidence interval compared to:

      Code:
      xtivreg2 , fe robust
      PS: I would appreciate it if Mark Schaffer or someone from Stata could shed light on the differences in their VCE estimators and the underlying assumptions.

      PSS: The following discussion between Ariel Karlinsky and Emad Shehata seems relevant: http://www.statalist.org/forums/foru...ivreg-xtivreg2 ... Emad Shehata suggestion to use "small" still does not estimate the same SE (a point also noted by Emad Shehata) in the same thread ... finally, Ariel Karlinsky seems to imply without further references that xtivreg2 may have more accurate SE estimates
      Did anyone finally answer this? I have the same issue exporting the complete outcome to LaTeX.
      Thanks!

      Comment

      Working...
      X