Announcement

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

  • Flexible polynomial regression?


    //I'm new here, sorry if the question is stupid and/or badly specified :/ Please help me what should I clarify

    I'm trying to do a variation of bunching (manipulating with wage levels to avoid higher tax rate/exam scores to pass etc) estimation by flexible polynomial OLS regression based on Kleven and Waseem 2013 (the model in question is on p. 21-23). To investigate the extent of bunching around the thresholds they first fitted a flexible polynomial (y=b0+b1*x+b2^2+b3^3+b4^4+b5^5...) to the empirical density of wage/points that excluded observations from the area around the threshold (that's a counter-factual) and then they repeated it, this time including the suspicious area (with dummy variables).

    Here's a picture to get the idea density comparison Here's the equation and a description the regression equation + model description

    I'm supposed to do that in Stata and I haven't been able to find a way how to make a flexible polynom there.

    Does anybody have an idea by any chance please?

  • #2
    You'll increase your chances of a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output (fixed spacing fonts help), and data using dataex.
    y=b0+b1*x+b2^2+b3^3+b4^4+b5^5.. This seems wrong. Why aren't the polynomials in the x's rather than the betas? I can't see that what you've written is even identifiable.

    Assuming your polynomial is in the x's, you can use factor variable notation to create the square, cube, etc. and then run ols.

    Comment

    Working...
    X