graph stata HR and tertiles

[cheng li yuang]

Good morning,
I'm trying to make a graph as follows after a cox model but I have never used stata before.

xi:xtmixed i.tertile_smoke

my outcome is mortality

I guess it's very simple for someone who knows how to use stata and graphs.

Any help?

Thank you
Cheng

[Joseph Coveney]

Welcome to Statalist. I think that if you're interested in fitting survival models with the Cox regression technique, then stcox would be a better choice than xtmixed. The graph comes easily afterward with the postestimation commands margins and marginsplot. As an illustration, run the following do-file and study the commands and output. (Focus on the commands after the "Begin here" comment; the first part is just to create a fictitious dataset for illustration.)

version 15.1

clear *

set seed `=strreverse("1489595")'
quietly set obs 100

generate double stress = runiform()
generate double longevity = -ln(runiform()) / 1 / exp(stress)
generate byte failure = 1

*
* Begin here
*
// Fitit
egen byte quartile = cut(stress), group(4)
label define Quartiles 0 First 1 Second 2 Third 3 Fourth
label values quartile Quartiles

stset longevity, fail(failure)

stcox i.quartile, nolog

// Graphit
margins quartile, predict(hr)

marginsplot , title("") ///
    plotopts(lcolor(black) mcolor(black) mfcolor(white)) ///
    ciopts(lcolor(black)) level(50) ///
    ylabel(0(1)3, angle(horizontal) nogrid) ///
    xtitle(Quartile of Stress) ///
    yline(1, lcolor(black) lpattern(dash))

exit

Be sure to look at the online help files and user's manual entries for the relevant commands that are used above.

help st
help stset
help stcox
help factor variables
help stcox_postestimation
help margins
help marginsplot

[cheng li yuang]

Joseph Coveney

Thank you so much +++ for your valuable help! I just wanted to ask something else. Actually, I just want to represent the HR and 95% confidence interval that are given in the stata output for my quartiles

xi:stcox i.quartile_smoke age sex ...
I'd like to plot the HR of the fist, 2nd and 3rd and 4th quartile with the 95%CI of the output in stata and margins doesn't allow to do that. I can only have predictions.

Could you help me with that? Otherwise thanks +++ for your help with the graph

Thank you again so much +++

Cheng LY

[JIANBO ZHOU]

HI Joseph Coveney

thanks for your help.

i met the similar question.

for example:

when some covariates are included in the stcox model,

code:

stcox age sex diabetes i.quartile, nolog
// Graphit margins quartile, predict(hr)
marginsplot , title("") /// plotopts(lcolor(black) mcolor(black) mfcolor(white)) /// ciopts(lcolor(black)) level(50) /// ylabel(0(1)3, angle(horizontal) nogrid) /// xtitle(Quartile of Stress) /// yline(1, lcolor(black) lpattern(dash))
exit but i could not get the right graph. could you help me with that? thank you! JIANBO

[JIANBO ZHOU]

Joseph Coveney