Combined graph in stata

[Marco Fioramanti]

Hello, I would like to draw a combined graph in which the contributions of the components to the total variation of the sum variable appear on the stacked bars, together with the line that describes the sum variable (see the attached graph in excel). How can I do it? Thank you in advance. Marco

[Chen Samulsion]

I am not before my pc, the following code maybe need to do some adjustment.

twoway bar dlpft capva lavva year, stack || connect diva year, yaxis(2) ylabel(-5(1)5)

[Nick Cox]

There is no data example here, and in any case the graph from MS Excel (which is what I think you mean) doesn't make much sense to me. Given components A, B. C which can be positive or negative (or zero) one recipe is

gen AB = A + B
gen ABC = AB + C

and then you can go

twoway bar A year || rbar A AB year || rbar AB ABC year

and ABC is automatically the resultant, which you can plot separately if you wish.

Almost none of these stacked bars start at zero, so let's guess that I am just misunderstanding what you want.

Independently of that, combining red and green on a display is not a good idea.

[Marco Fioramanti]

I am not before my pc, the following code maybe need to do some adjustment.

twoway bar dlpft capva lavva year, stack || connect diva year, yaxis(2) ylabel(-5(1)5)

This code would be the right code, but unfortunately "stack" option is not allowed with twoway.

[Marco Fioramanti]

There is no data example here, and in any case the graph from MS Excel (which is what I think you mean) doesn't make much sense to me. Given components A, B. C which can be positive or negative (or zero) one recipe is

gen AB = A + B
gen ABC = AB + C

and then you can go

twoway bar A year || rbar A AB year || rbar AB ABC year

and ABC is automatically the resultant, which you can plot separately if you wish.

Almost none of these stacked bars start at zero, so let's guess that I am just misunderstanding what you want.

Independently of that, combining red and green on a display is not a good idea.

I drew the excel graph very quickly, just to have a raw Idea. All the stacked bars start from 0, but on the rhs of the graph. The 0s of the 2 y axes are not aligned. The weird colours are just to have a clear idea. The graph represents the decomposition of value added (dlava) among its contributions) TFP, capital and labour. The right graph is the following (different scale).

[Nick Cox]

Thanks for explaining that the labels on the left in your graph are wrong. I still can't follow what the dashed line is doing.

There are more objections than stack being illegal to Chen's code as the bars aren't stacked even without it. So, it's not the code you seek.

My code is not right either as I guess you want negative components to be stacked separately from positive.

Nothing beats a good data example here!

[Marco Fioramanti]

Thanks for explaining that the labels on the left in your graph are wrong. I still can't follow what the dashed line is doing.

There are more objections than stack being illegal to Chen's code as the bars aren't stacked even without it. So, it's not the code you seek.

My code is not right either as I guess you want negative components to be stacked separately from positive.

Nothing beats a good data example here!

Below the data. dlva=dltfp+capva+lavva.

year	dlva	dltfp (dx)	capva (dx)	lavva (dx)
1996	0.012188	1.609203	-0.6817	-0.91531
1997	0.022875	0.630561	-0.64504	0.03735
1998	0.025538	2.005753	-0.58083	-1.39939
1999	0.054229	2.840631	-0.50738	-2.27902
2000	0.008153	4.304501	-0.30227	-3.99408
2001	-0.02877	-0.98376	-0.63488	1.58987
2002	-0.04827	1.721446	-0.599	-1.17071
2003	-0.03967	1.732056	-0.58573	-1.186
2004	0.110663	-0.33254	-0.50118	0.944377
2005	-0.0509	1.475137	-0.48415	-1.0419
2006	-0.00756	-0.32653	-0.35749	0.676464
2007	0.010813	1.235326	-0.47961	-0.7449
2008	0.020832	3.309675	-0.7994	-2.48945
2009	-0.01023	2.350233	-1.14812	-1.21235
2010	0.002476	1.591856	-1.01646	-0.57292
2011	0.007717	2.662456	-0.90211	-1.75263
2012	-0.04045	2.962323	-1.12566	-1.87712
2013	0.012591	1.70807	-1.55302	-0.14246
2014	-0.02155	0.556944	-1.51484	0.936341
2015	0.046213	0.501076	-1.47827	1.023405
2016	-0.00329	-0.243	-1.17417	1.413884
2017	-0.04081	1.94506	-0.84066	-1.14521
2018	0.02883	-0.95392	-0.57466	1.557407
2019	-0.02028	1.127075	-0.4755	-0.67185
2020	-0.0426	3.314502	-0.70852	-2.64859
2021	-0.00276	-1.62979	-0.29625	1.923284
2022	0.026298	-0.0801	-0.4453	0.551703
2023	-0.0359	2.108601	-0.48639	-1.65812

[Chen Samulsion]

"stack" option is not allowed with twoway

Sorry for that. You don't provide data example here. Let's use sp500. I recommend to use an user-written command -genstack- (ssc install genstack) to prepare the data. This command is designed to generate stacked variables for twoway stacked chart. However, I still don't know whether the following is what you want. By the way, Nick Cox's outdated command -barplot- is very interesting in this case. Below is the code.

ssc install genstack
ssc install barplot

sysuse sp500
gen low_neg=-low
gen close_neg=-close
genstack open high low_neg close_neg , gen(C_)
twoway bar C_close_neg C_low_neg C_high C_open date || connecte volume date, yaxis(2)
barplot C_close_neg C_low_neg C_high C_open date

[Chen Samulsion]

Thank you for the data. Here is the code, and I think it will be helpful for you. And you still need to install -genstack- command.

* Example generated by -dataex-. To install: ssc install dataex
clear
input float(year dlva dltfp capva lavva DLVA_dltfp DLVA_capva DLVA_lavva)
1996 .012188 1.609203   -.6817  -.91531 1.609203     -.6817   -1.59701
1997 .022875  .630561  -.64504   .03735  .630561    -.64504    .667911
1998 .025538 2.005753  -.58083 -1.39939 2.005753    -.58083   -1.98022
1999 .054229 2.840631  -.50738 -2.27902 2.840631    -.50738    -2.7864
2000 .008153 4.304501  -.30227 -3.99408 4.304501    -.30227   -4.29635
2001 -.02877  -.98376  -.63488  1.58987  -.98376   -1.61864    1.58987
2002 -.04827 1.721446    -.599 -1.17071 1.721446      -.599   -1.76971
2003 -.03967 1.732056  -.58573   -1.186 1.732056    -.58573   -1.77173
2004 .110663  -.33254  -.50118  .944377  -.33254    -.83372    .944377
2005  -.0509 1.475137  -.48415  -1.0419 1.475137    -.48415   -1.52605
2006 -.00756  -.32653  -.35749  .676464  -.32653    -.68402    .676464
2007 .010813 1.235326  -.47961   -.7449 1.235326    -.47961   -1.22451
2008 .020832 3.309675   -.7994 -2.48945 3.309675     -.7994   -3.28885
2009 -.01023 2.350233 -1.14812 -1.21235 2.350233   -1.14812   -2.36047
2010 .002476 1.591856 -1.01646  -.57292 1.591856   -1.01646   -1.58938
2011 .007717 2.662456  -.90211 -1.75263 2.662456    -.90211   -2.65474
2012 -.04045 2.962323 -1.12566 -1.87712 2.962323   -1.12566   -3.00278
2013 .012591  1.70807 -1.55302  -.14246  1.70807   -1.55302   -1.69548
2014 -.02155  .556944 -1.51484  .936341  .556944   -1.51484   1.493285
2015 .046213  .501076 -1.47827 1.023405  .501076   -1.47827   1.524481
2016 -.00329    -.243 -1.17417 1.413884    -.243   -1.41717   1.413884
2017 -.04081  1.94506  -.84066 -1.14521  1.94506    -.84066   -1.98587
2018  .02883  -.95392  -.57466 1.557407  -.95392   -1.52858   1.557407
2019 -.02028 1.127075   -.4755  -.67185 1.127075     -.4755 -1.1473501
2020  -.0426 3.314502  -.70852 -2.64859 3.314502    -.70852   -3.35711
2021 -.00276 -1.62979  -.29625 1.923284 -1.62979   -1.92604   1.923284
2022 .026298   -.0801   -.4453  .551703   -.0801 -.52540004    .551703
2023  -.0359 2.108601  -.48639 -1.65812 2.108601    -.48639   -2.14451
end

genstack dltfp capva lavva, gen(DLVA_)
twoway bar DLVA_lavva DLVA_capva DLVA_dltfp year, barwidth(0.7 0.7 0.7) || connected dlva year, yaxis(2) xlabel(1996(2)2022) ytitle("", axis(2)) legend(pos(6) row(1))

[Nick Cox]

Thanks for the data example. Meanwhile I was hacking away at this.

* Example generated by -dataex-. For more info, type help dataex
clear
input float(year A B C)
1 -1  5 1
2 -2  4 2
3  3  3 3
4  4 -2 2
5  5 -1 1
end

gen zero = 0 
gen posA = max(A, 0)
gen posAB = max(A, 0) + max(B, 0)
gen posABC = max(A, 0) + max(B, 0) + max(C, 0)
gen negA = min(A, 0)
gen negAB = min(A, 0) + min(B, 0)
gen negABC = min(A, 0) + min(B, 0) + min(C, 0)
gen total = A + B + C 

twoway bar posA year, col(stc1) barw(0.8) /// 
|| rbar posA posAB year, col(stc2) barw(0.8) /// 
|| rbar posABC posAB year, col(stc3) barw(0.8) /// 
|| bar negA year, col(stc1) barw(0.8) /// 
|| rbar negA negAB year, col(stc2) barw(0.8) /// 
|| rbar negABC negAB year, col(stc3) barw(0.8) ///
|| line total year, lc(black) lw(medthick) legend(order(1 "A" 2 "B" 3 "C" 7 "Total"))

[Chen Samulsion]

If you want to reproduce what you get in Excel, you need some adjust on the code above. For example, set xtitle as none (year is obvious and thus unnecessary), exchange the yaxis (use yaxis(2) to display the stacked bar), and lastly, rename the variables with stubname of DLAV_, etc.

Edit: here is my adjusted code.

rename (dltfp capva lavva) =_org
genstack dltfp_org capva_org lavva_org, gen(c_)
rename c_*_org *
twoway bar lavva capva dltfp year, barwidth(0.7 0.7 0.7) yaxis(2) || connected dlva year, yaxis(1) xlabel(1996(2)2022) xtitle("") ytitle("", axis(1)) legend(pos(6) row(1))

[Nick Cox]

Here's another take. I am not a great fan of stacked bars in general and not at all when magnitudes differ so much (and also in sign).

I used multiline from SSC.
https://www.statalist.org/forums/for...ailable-on-ssc

The data example in #7 won't run as code without minor surgery.

clear 
input year    dlva    dltfp     capva     lavva 
1996    0.012188    1.609203    -0.6817    -0.91531
1997    0.022875    0.630561    -0.64504    0.03735
1998    0.025538    2.005753    -0.58083    -1.39939
1999    0.054229    2.840631    -0.50738    -2.27902
2000    0.008153    4.304501    -0.30227    -3.99408
2001    -0.02877    -0.98376    -0.63488    1.58987
2002    -0.04827    1.721446    -0.599    -1.17071
2003    -0.03967    1.732056    -0.58573    -1.186
2004    0.110663    -0.33254    -0.50118    0.944377
2005    -0.0509    1.475137    -0.48415    -1.0419
2006    -0.00756    -0.32653    -0.35749    0.676464
2007    0.010813    1.235326    -0.47961    -0.7449
2008    0.020832    3.309675    -0.7994    -2.48945
2009    -0.01023    2.350233    -1.14812    -1.21235
2010    0.002476    1.591856    -1.01646    -0.57292
2011    0.007717    2.662456    -0.90211    -1.75263
2012    -0.04045    2.962323    -1.12566    -1.87712
2013    0.012591    1.70807    -1.55302    -0.14246
2014    -0.02155    0.556944    -1.51484    0.936341
2015    0.046213    0.501076    -1.47827    1.023405
2016    -0.00329    -0.243    -1.17417    1.413884
2017    -0.04081    1.94506    -0.84066    -1.14521
2018    0.02883    -0.95392    -0.57466    1.557407
2019    -0.02028    1.127075    -0.4755    -0.67185
2020    -0.0426    3.314502    -0.70852    -2.64859
2021    -0.00276    -1.62979    -0.29625    1.923284
2022    0.026298    -0.0801    -0.4453    0.551703
2023    -0.0359    2.108601    -0.48639    -1.65812
end 

multiline dlva    dltfp     capva     lavva year, xla(1996(4)2020 2023) yli(0) xtitle("")

[Marco Fioramanti]

If you want to reproduce what you get in Excel, you need some adjust on the code above. For example, set xtitle as none (year is obvious and thus unnecessary), exchange the yaxis (use yaxis(2) to display the stacked bar), and lastly, rename the variables with stubname of DLAV_, etc.

Edit: here is my adjusted code.

rename (dltfp capva lavva) =_org
genstack dltfp_org capva_org lavva_org, gen(c_)
rename c_*_org *
twoway bar lavva capva dltfp year, barwidth(0.7 0.7 0.7) yaxis(2) || connected dlva year, yaxis(1) xlabel(1996(2)2022) xtitle("") ytitle("", axis(1)) legend(pos(6) row(1))

The graph you drew above is what I want, but when I try to reproduce I get all the points connected:

[Chen Samulsion]

The graph you drew above is what I want, but when I try to reproduce I get all the points connected

Well, just add sort option after twoway connected. Here it is:

rename (dltfp capva lavva) =_org
genstack dltfp_org capva_org lavva_org, gen(c_)
rename c_*_org *
twoway bar lavva capva dltfp year, barwidth(0.7 0.7 0.7) yaxis(2) || connected dlva year, sort yaxis(1) xlabel(1996(2)2022) xtitle("") ytitle("", axis(1)) legend(pos(6) row(1))

[Marco Fioramanti]

Here's another take. I am not a great fan of stacked bars in general and not at all when magnitudes differ so much (and also in sign).

I used multiline from SSC.
https://www.statalist.org/forums/for...ailable-on-ssc

The data example in #7 won't run as code without minor surgery.

clear
input year dlva dltfp capva lavva
1996 0.012188 1.609203 -0.6817 -0.91531
1997 0.022875 0.630561 -0.64504 0.03735
1998 0.025538 2.005753 -0.58083 -1.39939
1999 0.054229 2.840631 -0.50738 -2.27902
2000 0.008153 4.304501 -0.30227 -3.99408
2001 -0.02877 -0.98376 -0.63488 1.58987
2002 -0.04827 1.721446 -0.599 -1.17071
2003 -0.03967 1.732056 -0.58573 -1.186
2004 0.110663 -0.33254 -0.50118 0.944377
2005 -0.0509 1.475137 -0.48415 -1.0419
2006 -0.00756 -0.32653 -0.35749 0.676464
2007 0.010813 1.235326 -0.47961 -0.7449
2008 0.020832 3.309675 -0.7994 -2.48945
2009 -0.01023 2.350233 -1.14812 -1.21235
2010 0.002476 1.591856 -1.01646 -0.57292
2011 0.007717 2.662456 -0.90211 -1.75263
2012 -0.04045 2.962323 -1.12566 -1.87712
2013 0.012591 1.70807 -1.55302 -0.14246
2014 -0.02155 0.556944 -1.51484 0.936341
2015 0.046213 0.501076 -1.47827 1.023405
2016 -0.00329 -0.243 -1.17417 1.413884
2017 -0.04081 1.94506 -0.84066 -1.14521
2018 0.02883 -0.95392 -0.57466 1.557407
2019 -0.02028 1.127075 -0.4755 -0.67185
2020 -0.0426 3.314502 -0.70852 -2.64859
2021 -0.00276 -1.62979 -0.29625 1.923284
2022 0.026298 -0.0801 -0.4453 0.551703
2023 -0.0359 2.108601 -0.48639 -1.65812
end

multiline dlva dltfp capva lavva year, xla(1996(4)2020 2023) yli(0) xtitle("")

[ATTACH=CONFIG]n1771580[/ATTACH]

My fault. There was a bug in the code generating variables so that very different scale.