Announcement

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

  • Heteroskedastic errors and svy

    Dear All,

    Svy: reg is not letting me perform -hettest- or -imtest- postestimation.

    Is the rationale behind this that errors are already corrected for heteroskedasticity when I'm using the svy prefix?

    Does anyone know any other way of checking for heteroskedasticity?

    Thanks in advance.

  • #2
    Alina:
    you may want to consider visual inspection.
    Kind regards,
    Carlo
    (Stata 18.0 SE)

    Comment


    • #3
      As you surmised, svy does trigger variance estimation robust to heteroskedasticity. See the paragraph preceding Example 4 in the introduction to survey commands.
      Steve Samuels
      Statistical Consulting
      [email protected]

      Stata 14.2

      Comment


      • #4
        Originally posted by Carlo Lazzaro View Post
        Alina:
        you may want to consider visual inspection.
        Thanks Mr Lazzaro. Will you please share the commands/procedure to do so?

        Comment


        • #5
          Originally posted by Steve Samuels View Post
          As you surmised, svy does trigger variance estimation robust to heteroskedasticity. See the paragraph preceding Example 4 in the introduction to survey commands.
          Thanks a lot for the reference! Cleared up my confusion.

          Comment


          • #6
            Alina:
            see -help rvfplot-.
            Kind regards,
            Carlo
            (Stata 18.0 SE)

            Comment


            • #7
              Originally posted by Carlo Lazzaro View Post
              Alina:
              see -help rvfplot-.
              Thanks a lot!

              Comment


              • #8
                Hello,
                I am finding this tread useful since I have the very same question as of the starter.
                Steve Samuels, thanks for the insight, but I could not find any "Example 4" from inside link you have sent as well as could not find any mention of "hetero/homoskedasticity" or "robust*" by searching by keywords.
                Could you please be more specific about the reference?
                Thank you in advance!

                Best,
                Linda

                Comment


                • #9
                  Linda:
                  I guess Steve meant Example 3 under -survey- entry, Stata .pdf manual.
                  Kind regards,
                  Carlo
                  (Stata 18.0 SE)

                  Comment


                  • #10
                    Thank you Carlo!

                    1) Again, maybe it is just me being helpless but before Example 3 I can only see Example 2 and there they talk about the fact that ignoring stratification and PSU cause inefficient standard errors. Maybe that has to do with heteroskedasticity correction?

                    2) Carlo: I have tried -rvfplot-. after running my regression - reg - (Note: I have fed in weights bit not PSU nor stratification since those are not available in my dataset).
                    I get the error message in red.
                    2.1. What is wrong?
                    2.2. How can check for heteroskedasticity in some other way?
                    2.3.Does svy still fix it if I don't feed in PSU and stratification)





                    svyset [pweight=weighta]
                    Click image for larger version

Name:	vv.PNG
Views:	1
Size:	33.9 KB
ID:	1602828

                    Last edited by Linda Luciani; 11 Apr 2021, 04:18.

                    Comment


                    • #11
                      Just found a good answer to my enquires at #2 of this tread (link ), if anyone is curious, although no textbook, stata pdf for reference is provided.
                      Best,
                      Linda

                      Comment


                      • #12
                        Linda:
                        you did nothing wrong: -rvplot- is simply not allowed after -svy: regress-.
                        Thankd for sharing the Stata thread that helped you out.
                        Something similar is reported in -survey- entry, Stata .pdf manual, Survey data analysis tools, page 6.
                        Kind regards,
                        Carlo
                        (Stata 18.0 SE)

                        Comment


                        • #13
                          Amazing Carlo!
                          No Idea where I have been looking before...now I could clearly spot the reference!

                          Thank you very much for your help!

                          Best,
                          Linda

                          Comment


                          • #14

                            My comment in 2018 was misleading. I apologize.

                            . The paragraph that precedes example 4 , "Survey data analysis tools" , says only that the linearization variances are "robust", but doesn't say what they are robust to. They are robust to heteroskedasticity, clustering, and survey design( if the data are properly svyset).

                            Stata's svy replication methods -brr, sdr, jackknife- are also "robust",
                            Steve Samuels
                            Statistical Consulting
                            [email protected]

                            Stata 14.2

                            Comment

                            Working...
                            X