Deriving vertex co-ordinates from a qfit plot

[Jay Tanner]

Hi everyone,

I'm new to Stata so please forgive me if I get any of my terms wrong. I've generated qfit plots to visualise the relationship between my dependent and independent variables (two-way qfit y x) but now i want to determine the co-ordinates of the vertex. I can do this manually by regressing the variables (regress y x x^2) to generate coefficients and then calculate the x value for the vertex using '-linear coefficient/(2xquadratic coefficient)' and lastly calculate the y value by inserting the x value back into the regression derived quadratic equations. However, i have 20 qfits so this will take hours. Is there any shortcut for this in stata?

Many thanks

Jay

Edit- I'm using small stata v14.2

[Mike Lacy]

A loop is in order here, the exact form of which would depend on the nature of your list of x and y variables and possibly what you want to do with these results after you get them. Here's an illustration of one way to do this. I didn't check the results at all carefully, so you will want to check some examples by hand:

sysuse auto
local y = price
foreach x of varlist weight displacement length {
   regress price c.`x'##c.`x'
   local b1 = el(r(table),1,1)
   local b2 = el(r(table),1,2)
   local a = el(r(table),1,3)
   local xval = -`b1'/(2*`b2')
   local yval = `a' + `b1' * `xval' + `b2' * `xval'^2
   display "xvar = `x'; x, y = " `xval' ", " `yval'
}

[Jay Tanner]

Thank you very much!