floatplot -legend- option

[Marc Kaulisch]

Dear Nick Cox , I have started using your -floatplot- from SSC program. I really like to use it more extensively. But I wonder about the legend-options behaviour.
This is what I see in your ado (-vertical-option not used):

 legend(order(`legend') row(1) pos(6))

Is there a good reason about this setting?

If I want to tweak the -legend-options I get not reliable results. E.g. if I add

legend(col(1) pos(5))

to the -floatplot- command I get the legend at "row(1) pos(6)" and not my change.

* Example generated by -dataex-. For more info, type help dataex
clear
input byte(jahr finr) double fi_
4 1 2296399
4 2  115000
4 3 1488100
4 4  484899
4 5  459300
4 6  423200
4 7  132200
4 8  595900
3 1 2002099
3 2  108300
3 3 1483100
3 4  406399
3 5  341900
3 6  356699
3 7  136300
3 8  694099
2 1 1743699
2 2   78000
2 3 1252200
2 4  357399
2 5  288700
2 6  303099
2 7  115600
2 8  586099
5 1 2333000
5 2   77500
5 3 1698299
5 4  480500
5 5  417900
5 6  493900
5 7  203800
5 8  729200
6 1 2971799
6 2   60600
6 3 2051000
6 4  549300
6 5  439899
6 6  596700
6 7  153099
6 8  610900
1 1       0
1 2       0
1 3       0
1 4       0
1 5       0
1 6       0
1 7       0
1 8       0
end
label values jahr jahr
label def jahr 1 "2024", modify
label def jahr 2 "2057", modify
label def jahr 3 "2080", modify
label def jahr 4 "2113", modify
label def jahr 5 "2146", modify
label def jahr 6 "2179", modify
label values finr l_finr
label def l_finr 1 "Ein", modify
label def l_finr 2 "Foz", modify
label def l_finr 3 "So", modify
label def l_finr 4 "Schw", modify
label def l_finr 5 "Fo", modify
label def l_finr 6 "Gra", modify
label def l_finr 7 "Gr", modify
label def l_finr 8 "Ex", modify

My command is based on the second example in the helpfile:

floatplot finr [fw=fi_], centre(3) fcolors(gs6 green orange bluishgray maroon blue yellow purple)  over(jahr) note("% zentriert auf So" "Quelle:", span) name(float_fi, replace) legend(col(1) pos(5)) xtitle("") ytitle("")

[Marc Kaulisch]

Okay, I tweaked legend-options to

legend(col(1) row(8) pos(4))

and that works for me at the moment.
So I learned that I need to use the legend-row-option in order to tweak the legend when in the non-vertical-option mode.

[Nick Cox]

Good that you solved your problem.

The default for floatplot legends is just what is and probably grows out of what worked well for my own examples. Oddly, or otherwise, I can't find a thread earlier this year announcing floatplot so did you learn about it from Statalist or from some talk I gave this summer?

Good too if you like the result. The data appear to be time series. so that suggests alternatives. Here is one experiment:

gen outcome = fi_/1e6
label var outcome "whatever (millions)"
decode jahr, gen(year)
destring year, replace
levelsof year
local levels `r(levels)'

// fabplot is from the Stata Journal 
fabplot line outcome year, by(finr, col(2)) xla(`levels', labsize(small)) front(connected) frontopts(lwidth(*2)) xtitle("")

Naturally there is room for less cryptic labels.

Sometimes a table works as well!

[Marc Kaulisch]

Dear Nick, yeah, I am not sure where I heard about -floatplot-. It may also be when I wanted to learn about what's new on ssc with -ssc new-.. Today I found one other thread in which -floatplot- is mentioned.
Indeed the data is time series. Thanks for this suggestion. It is another plot I wanted to test the coming days.
For the question I have at the data I think -floatplot- is really good because it shows how the importance of a category within an unit changes over time. Whereas -fabplot- - in my interpretation- highlights the development of each category by itself contrasted by the development of the other categories.
Cryptic labels used to hide original data.