Use lgraph to draw a parellel trend graph.

[Jacky Lin]

Hi,I am a new Stata user. I am using the hospdd.dta to run a DID program. I want to use the code lgraph to generates the parellel trend.

Here is my data

I use the code: lgraph satis month, by(procedure) to generate the following graph:

But the correct graph should be like this, it seems that the first three months are disappear for the control group.

What should be the correct code? Really appreciate your answers!

[Jacky Lin]

I also following this code to generate the parellel trend graph:

bysort month procedure: egen mean_satis = mean(satis)

twoway (connected mean_satis month if procedure==0, msize(medium) msymbol(circle) lwidth(thin) lpattern(dash)) (connected mean_satis month if procedure==1, msize(medium) msymbol(circle) lwidth(thin) lpattern(solid)), graphregion(color(white)) xline(3,lc(red) lpattern(solid)) ytitle(Patient satisfaction score, size(medium)) xlabel(1(1)7,labsize(medium)) ylabel(3.4(0.2)4.4,labsize(medium)) legend(label(1 "Control") label(2 "Treatment") size(medium))

And get this graph:

I also can not plotted the treatment group from the first three month.

[Jared Greathouse]

I don't want to see a picture, please use dataex as the FAQ says to show us your real dataset, one that I or someone else can actually work with.

[Roman Mostazir]

There has not been a single occurrence in your data picture! where procedure=1 and month is less than 4. Which is the reason your 'treatment' (procedure=1) group has a line starting from month=4. What is the data source of the claimed 'correct' graph? Are both graphs being generated from the same data? I don't think so. Show the data example again using -dataex- ideally with a blend of procedure and month so that we can see what is happening here. Please don't use picture for data example, use -dataex- as suggested in #3

[Enrique Pinzon (StataCorp)]

Hi Jacky,

I would try the following code

webuse hospdd, clear
bysort hospital: egen treated = max(procedure)
lgraph satis month, by(treated)

Jared and Roman give you good advice. To try to get better help, providing reproducible examples, data and code, is important. I just happen to be a bit familiar with the data you are mentioning. The key here is that the variable procedure is the interaction of the treated group and the post-intervention period. You just need to define a treatment group to reproduce the graph. Welcome to Stata and to Statalist.

[Jacky Lin]

I don't want to see a picture, please use dataex as the FAQ says to show us your real dataset, one that I or someone else can actually work with.

Hi Jared, Thank you so much! Now I know how to post the data.

[Jacky Lin]

There has not been a single occurrence in your data picture! where procedure=1 and month is less than 4. Which is the reason your 'treatment' (procedure=1) group has a line starting from month=4. What is the data source of the claimed 'correct' graph? Are both graphs being generated from the same data? I don't think so. Show the data example again using -dataex- ideally with a blend of procedure and month so that we can see what is happening here. Please don't use picture for data example, use -dataex- as suggested in #3

Hi Roman, Thank you for the suggestion!

[Jacky Lin]

Hi Jacky,

I would try the following code

webuse hospdd, clear
bysort hospital: egen treated = max(procedure)
lgraph satis month, by(treated)

Jared and Roman give you good advice. To try to get better help, providing reproducible examples, data and code, is important. I just happen to be a bit familiar with the data you are mentioning. The key here is that the variable procedure is the interaction of the treated group and the post-intervention period. You just need to define a treatment group to reproduce the graph. Welcome to Stata and to Statalist.

Hi Enrique, Thank you so much for your code! I can generate the same graph as yours. Really appreciate it!