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

  • -lfitci- and narrowing the scale of an axis

    range() never narrows the scale of an axis or causes data to be omitted from the plot.
    This is a principled decision by StataCorp. Nevertheless, it causes me headaches. The headache I'm facing right now is that graphs not only expand to include all data, but also expand to include non-data, such as confidence intervals drawn with -lfitci-.

    As a contrived example:

    sysuse nlsw88
    graph twoway lfitci wage tenure if industry == 2
    Stata happily devotes a third of the graph to negative-wage territory, since the lower bound of the CI extends into it. I don't suppose anybody has a neat way of addressing that? (A general solution, not just dropping the entire lower bound of estimation.)