Calculating turning point for moderated quadratic effects

[Matt Piszczek]

Dear Forumgoers,
I have a mixed effects model in which individuals (designated by "id") are measured on X and Y five days in a row which includes a quadratic term for X. I am interested in calculating the X value for the turning point at each of the five day values. The model includes random intercepts and slopes.
I was able to calculate the turning point for the initial model before adding the interaction term as follows:

 mixed y c.x##c.x i.day || id: c.x
 nlcom -_b[c.x]/(2*_b[c.x#c.x])

However, I am not sure how to adjust this code to get the predicted turning points for each value of "day" (which ranges 1 to 5) when I add the interaction as follows.

 mixed y c.x##c.x##i.day || id: c.x

It seems like I can adjust the nlcom prediction, but I am unsure how. Can anyone clarify?

Thanks,
Matt

[Clyde Schechter]

Not tested, so beware of typos, unbalanced or misplaced parentheses, etc.:

forvalues d = 1/5 {
    nlcom -(_b[c.x]+_b[`d'.day#c.x])/(2*(_b[c.x#c.x]+_b[`d'.day#c.x#c.x]))
}

[Matt Piszczek]

Thanks Clyde--this worked perfectly and I see exactly how it works. Your help is much appreciated.