[Marco Fioramanti]

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))

[ATTACH=CONFIG]n1771586[/ATTACH]

Not yet there. I get dots for every variable and in addition stacked bars do not start from 0.

[Nick Cox]

I guess that #13 also means that your data are more complicated than the data example you gave.

[Chen Samulsion]

I agree with Nick. It seems that there are more than one observations in each year. You should check your data, and grasp its structure.

[Marco Fioramanti]

Nick's #17 hint was a great one: in fact I forgot (because of excel) I have a panel.
Now I have the right one. Sorry for the many misleading inaccuracies and thank you very much for your valuable help.

[Nick Cox]

I am glad you like the graph you got, but different preferences aside I have to suggest that stc4 is a poor colour for dlva -- supposedly, or presumably, the most important variable. I recommend black and a thicker line, as in an earlier post, #10.

[Marco Fioramanti]

I am glad you like the graph you got, but different preferences aside I have to suggest that stc4 is a poor colour for dlva -- supposedly, or presumably, the most important variable. I recommend black and a thicker line, as in an earlier post, #10.

Dear Nick,
As I told you that was just a draft of the graph whose goal was to understand if that kind of graph was possible in stata. In fact, the "genstack" user written program is the key. Now I have to conform to my Institute (in addition to carefully review data, scale, axes and so on) for all the colours: black for lines or connected lines, different blue nuances for bars. But this is a minor problem I guess.
Thank you again for your valuable help.
Best regards,
Marco

[Nick Cox]

We're at cross purposes.

I am comparing the colours in #19 with those in #10, not reverting to discussion of #1. Naturally, your house style now takes over.

But since you've mentioned it, the use of genstack is strictly optional here. You just need to keep track of positive and negative values separately. My code in #3 failed there but I think #10 nailed it.