Announcement

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

  • Most suitable regression model

    Hello,
    I have a dependent variable which is ranked from 0 to 1, continuous and left-skewed (see attached image). I have almost 1,000 observations. Which regression model might be best? Tobit? Someone else mentioned quasi-binominal, but this isn't on Stata. Any advice would be greatly appreciated,
    Tom


    Click image for larger version

Name:	variabledistribution.png
Views:	1
Size:	90.7 KB
ID:	1718436

  • #2
    Quasi-binomial [not binominal] is available through glm, link(logit) vce(robust) f(binomial) -- as discussed in (e.g.) http://www.stata-journal.com/article...article=st0147

    See also
    betareg.

    Comment


    • #3
      Thanks Nick - that's really useful :-)

      Comment


      • #4
        Dear Nick, I typed glm, link(logit) vce(robust) f(binomial)into Stata but got 'last estimates not found'? If you have a minute, your thoughts would be greatly appreciated,
        Tom

        Comment


        • #5
          You do need to indicate the variables that the command should use. See help glm.

          Comment


          • #6
            Thanks so much for your help! I'm looking at it now :-)

            Comment

            Working...
            X