overlap of bars using rbar

[Federica Di Marcantonio]

Dears
I am trying to place side by side box for the year 2016 and 2006. I was not able to use by with rbar and with the following command the two sets of information overlap. Is there a way to put the two boxes side by side? I copy the data I used and the code.

Thank you in advance

 twoway   rbar lqt med Reg, pstyle(p1) barw(.6) || ///
       rbar med uqt Reg, pstyle(p1) barw(.6) || ///
       rspike lqt ls Reg, pstyle(p1) || ///
       rspike uqt us Reg, pstyle(p1) || ///
       rcap ls ls Reg, msize(*6) pstyle(p1) || ///
       rcap us us Reg, msize(*6) pstyle(p1) || ///
       scatter highest Reg, msymbol(Oh) msize(*2)  || ///
    scatter lowest Reg, msymbol(Oh) msize(*2)   ///
       xlabel( 1 "A" 2 "B" 3 "C" 4 "D" 5 "E") ytitle(price)

* Example generated by -dataex-. To install: ssc install dataex
clear
input float(Regions med lqt uqt ls us highest lowest year)
1   28 27 30  22.5 34.5 55 10 2016
2   32 30 35  22.5 42.5 65 24 2016
3   30 29 33    23   39 48 10 2016
4   27 26 30    20   36 57 19 2016
5   30 20 45 -17.5 82.5 86  0 2016
1   35 32 38    23   47 56 10 2006
2 36.5 33 40  22.5 50.5 70 21 2006
3   35 33 36  28.5 40.5 49 24 2006
4   36 34 39  26.5 46.5 49 22 2006
5   29 10 43 -39.5 92.5 90  0 2006
end
label values Regions reg
label def reg 1 "A", modify
label def reg 2 "B", modify
label def reg 3 "C", modify
label def reg 4 "D", modify
label def reg 5 "E", modify

[Nick Cox]

This is legal, but I can't tell whether it is what you want.

twoway   rbar lqt med Reg, pstyle(p1) barw(.6) || ///
       rbar med uqt Reg, pstyle(p1) barw(.6) || ///
       rspike lqt ls Reg, pstyle(p1) || ///
       rspike uqt us Reg, pstyle(p1) || ///
       rcap ls ls Reg, msize(*6) pstyle(p1) || ///
       rcap us us Reg, msize(*6) pstyle(p1) || ///
       scatter highest Reg, msymbol(Oh) msize(*2)  || ///
    scatter lowest Reg, msymbol(Oh) msize(*2)   ///
       xlabel( 1 "A" 2 "B" 3 "C" 4 "D" 5 "E") ytitle(price) by(year)

That code looks horribly reminiscent of http://www.stata-journal.com/sjpdf.h...iclenum=gr0039 (corrected by http://www.stata-journal.com/sjpdf.h...lenum=gr0039_1), "horribly" for me because if I want box plots I now usually go for stripplot (SSC).

With 5 regions and 2 years, I would

1. Tend to put years side by side, not what the code above does.

2. Tend to show more of the data.

Here is a reproducible example:

clear
set obs 100
set seed 2803

gen year = _n <= 50
label def year 0 2006 1 2016
label val year year

egen seq = seq(), block(10) to(5)
gen region = word("A B C D E", seq)

gen outcome = rnormal()

stripplot outcome, over(year) by(region, row(1) compact) centre vertical box

[Federica Di Marcantonio]

Dear Nick

thank you very much for your suggestion this is what I needed.

Notwithstanding I have another issue to solve. In my dataset I have two additional and separate variables that corresponds to the highest and lowest level in the year ( what I represent in the boxes are the value of a specific month). I would like to include these information as well in graph. I have tried the following but the dots are not in line with boxes. Is there a way to include these additional info?

stripplot myvar, over(year) by(Reg, row(1) compact) centre vertical box ///
addplot( ///
scatter highest_2016 Reg if year==1  || ///
scatter lowest_2016  Reg if year==1 || ///
scatter highest_2006 Reg if year==0 ||  ///
scatter lowest_2006 Reg if year==0 ///
 )

Thanks

Federica

[Nick Cox]

If you look at my example, the over() variable defines the x axis.

I guess you need therefore

stripplot myvar, over(year) by(Reg, row(1) compact) centre vertical box ///
addplot(scatter highest* lowest* year)

but without a data example I cannot be sure.

[Federica Di Marcantonio]

I did a small change because in each year it appeared the two different dots (that is in the year 2006 I had both the lowest/higher of 2006 and 2016). Basically this is the code

stripplot myvar over(year) by(Reg, row(1) compact) centre vertical box  ///
 addplot(scatter highest_2006 lowest_2006 year if year==0 || ///
 scatter highest_2016 lowest_2016 year if year==1)

Thank you very much for helping!!!

[Federica Di Marcantonio]

Dears,

I did a small change in the command above (post #5 ) but apparently I do something wrong considering that I get the following error message:

variable highest_2006 not found
r(111);

stripplot myvar1  myvar2, by(Reg, row(1) compact) centre vertical box  ///
 addplot (scatter highest_2006 lowest_2006 Reg  || ///
 scatter highest_2016 lowest_2016 Reg )

Can anyone tell me what the issue is with my current command?

Thank you in advance

[Nick Cox]

Is it true? That is, do you have a variable highest_2016 as you need?

Otherwise it would do no harm to close up

addplot (

to

addplot(

[Federica Di Marcantonio]

when I run the command

 
 scatter highest_2006 lowest_2006 Reg

it gives me the graph i need.
Apparently it is not a matter of space.

[Nick Cox]

OK. Looking again at

 
 stripplot myvar1  myvar2, by(Reg, row(1) compact) centre vertical box  ///  addplot(scatter highest_2006 lowest_2006 Reg  || ///  scatter highest_2016 lowest_2016 Reg )

it's not obvious to me what you're expecting. The stripplot is going to have an x axis that is devised for the purpose of showing myvar1 myvar2 distributions side by side -- and you also want that done separately by(Reg). So far, so good.

But you also want to superimpose scatter plots in the same space with Reg on the x axis.

I don't see that that would be illegal and don't see why you got that error message, but it is a puzzling specification.

Can you post a minimal data example that allows the error to be reproduced?