Announcement

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

  • How do I change the color scheme on the individual plots in an AV plot?

    Hi Folks,

    I'm generating a set of AV plots for all my explanatory variables using the following code:
    avplots, scheme(s1mono) mlabel(site_name) mlabp(3) mlabsize(small)


    Although the frame that holds the individual plots is now following the s1mono color scheme, the individual plots are still in Stata green (see attached image).

    Does anyone know the trick to fix this?

    Thanks!

    -nick
    Attached Files

  • #2
    I think there is a conflict between the individual plots which are drawn by avplot and the combination which is done by avplots. The scheme() option is grabbed by one command and so not seen by the other.

    If you

    Code:
     
    set scheme s1mono
    before the command then it should apply globally

    That said: A regression model with 11 data points and 5 predictors??? I guess you know that most researchers would want to flag that you are pushing hard there.

    Comment


    • #3
      Dear Dr. Cox,

      Thank you -- that code tweak solved the problem!

      As to your larger point: Yes, I realize that my model shouldn't work very well -- and that's why my dissertation chair is asking me to run these diagnostic plots! Despite the paltry degrees of freedom the p-value of the F test is acceptable (0.034) and the Adj R-squared is high: 0.7200.

      In addition, I've run VIF tests (all regressors below 3), a Hettest (p-value of 0.69) and a Wald Test for join significance (p-value 0.0004). Given the heteroskedasticity, I re-ran the model with Robust SE, none of the signs flip and the significant variables become more so. The F-test remains significant.

      Any idea why this model performs as well as it does in terms of R-squared? Are there additional tests I can use to ensure the results are not spurious?

      I'm attaching the outputs as images.

      Thanks for any insight!

      -nick
      Attached Files

      Comment


      • #4
        I'd try much simpler models; I don't make a fetish of P-values, but you are adding predictors whose effects are not significant at conventional levels.

        Comment

        Working...
        X