Twoway graph generates vertical lines at min/max values of predictor

[Matt Lenard]

Dear Statalist,

Why does my twoway graph generate vertical lines at the minimum and maximum values of my treatment variable (i.e., 0,1)? I would like to remove them.

Thanks so much,

Matt

Regression model:

 eststo: mixed math treatment math_prior covariates i.year || school: , covariance(unstructured) mle

Graph code:

predict values, fitted
sort school treatment
twoway (line values treatment), ///
graphregion(color(white)) ///
xtitle(" ""Treatment") ///
xscale(range(-.1 1.1)) ///
xlabel(0 1) ///
ytitle("Empirical Bayes Regression Lines, Math"" ")

Graph:

[Scott Merryman]

Because the lines at 0 and 1 is where the predicted values are for the treatment variable. The sloped lines are arbitrary line connecting changes in treatment from one school to another. If you had sorted:

sort treatment school

there would one connecting line.

For example:

webuse nlswork,clear
mixed ln_w grade age   collgrad || id:
predict value, fitted
sort id coll
twoway connect value collgrad, title(Sorted on id collgrad) name(gr1,replace)
sort coll id
twoway connect value collgrad, title(Sorted on collgrad id) name(gr2,replace)

[William Lisowski]

Perhaps adding the connect(L) option to the twoway command will solve your problem, or at least point you on your way to a solution.

[Matt Lenard]

Thanks Scott. That totally makes sense now. The graphs I was referencing Rabe-Hesketh & Skrondal had a continuous predictor variable.

[wbuchanan]

Matt Lenard
Hit me up sometime and I can show you a bunch of Stata related stuff that you might find useful. I wasn’t sure if there was another Matt Lenard out there. Hope things are going well at HGSE.