Frequency labels in a horizontal histogram

[Sabine Keller]

Hi,

I want to draw a histogram that meets two requirements:

1. displaying value labels readable
2. displaying the frequency of each category.

When a draw the histogram vertical, I don't have readable value labels. When I draw it horizontal the frequencies look very strange.

    qui sum V3_2
    local mw = r(mean)
    hist V3_2, addlabels xline(`mw', lwidth(vthick)) frequ /*percent*/ discrete xlabel(1 (1) 5, valuelabel angle(horizontal)) /*
        */ color(none) lwidth(medthick) lcolor(black)  /*
        */ xtitle("Häufigkeit") ytitle("")

    hist V3_2, addlabels yline(`mw', lwidth(vthick)) frequ /*percent*/ discrete ylabel(1 (1) 5, valuelabel angle(horizontal)) /*
        */ color(none) lwidth(medthick) lcolor(black)  /*
        */ xtitle("Häufigkeit") horizontal ytitle("")

Does anyone know how to modify the addlabels so that they look nice in a horizontal histogram?
Thanks in advance!

Kind regards
Sabine

[Sabine Keller]

Unfortunately I'm out of office during the next days. So please excuse me, if I don't respond immediately.

[Isaac Maddow-Zimet]

Hi Sabine,

I think the easiest way to do this would be to use -graph hbar- instead. Using the auto dataset, this would look something like:

sysuse auto, clear
graph hbar (count), over(foreign) blabel(total)

Where the variable foreign would be replaced with V3_2, in your case (which is a categorical variable, right?). You could then of course play with the other graph options to get at something closer to what you have above, style-wise.

Does this go some ways towards solving your problem?

Isaac

[Sabine Keller]

Hi Isaac,

Thanks for the hint. The frequency labels work but unfortunately I despair of the horizontal line xlines(1.846153846153846) not allowed, xaxis1 does not exist and to plot the whole scale (with blank space if no one chose this category): xlabels(1(1)4) not allowed, xaxis1 does not exist.

Kind regards
Sabine

[Nick Cox]

This may help.

sysuse auto, clear 
set scheme s1color 

preserve 

label def rep78 1 abysmal 2 appalling  3 adequate 4 appealing 5 amazing 
label val rep78 rep78 
su rep78, meanonly 
local mean = r(mean) 

contract rep78, nomiss 

twoway bar _freq rep78, horizontal yla(1/5, valuelabel ang(h) tlc(none))barw(0.8) bfcolor(none) /// 
|| scatter rep78 _freq, ms(none) mla(_freq) mlabpos(9) yli(`mean') ysc(titlegap(*5)) legend(off)  

restore

[Nick Cox]

Or see tabplot (SSC; SJ 17(3) and so downloadable with Stata 15 updated to 25 Sept 2017).

See e.g. https://www.statalist.org/forums/for...updated-on-ssc for an overview.

[Sabine Keller]

Hi Nick,

tabplot is really good! I like it! Thanks.
I have two small questions about it left:

1. Is it possible to rename the "percent" on top?
2. Is it possible to display percent and frequencies in the labels?

Thanks,
Sabine

[Nick Cox]

Glad you like it.

Both questions are already answered in the help. Make sure that you get the updated files as released 25 September, but the questions were answered before the latest revision.

#1 is a matter of the subtitle() option and just entails overwriting the default.

#2 There is a reproducible example showing how to do exactly that. Search for "radiologist assessment data" in the help.

[Sabine Keller]

Hi Nick,

The more I know, the more I don't know ...
The help file says: showval specifies that numeric values be shown beneath (or if horizontal is specified, to the left of) bars. Unfortunately the bar label is (with frequencies and percents) too long for the left side. Is there an opportunity to rearrange the bar label?

Regards
Sabine

[Nick Cox]

Please show example data and exact code. (FAQ Advice #12)

What does your longest label look like?

[Sabine Keller]

qui sum G1
local mw = r(mean)
qui inspect G1
local nm=r(N_pos)
cap drop show
bysort G1: gen show = string(_N) + " | " + string(_N*100/`nm', "%2.1f") + " %"
tabplot G1, horizontal showval(show) yline(`mw') ylabel(1 (1) 5) yasis xreverse percent  /*
    */ ytitle("") color(none) lwidth(medthick) lcolor(black) subtitle("Häufigkeit und Prozent")

[Nick Cox]

No example data there, but it allows a suggestion. You can get more space by using e.g.

xsc(r(0.8 .))

where 0.8 is the number to tune.

[Sabine Keller]

Perfect! Thanks a lot!