Change axis in graph bar

[Jean Jacques]

Hi Statausers,

I think is a common issue, but I couldn't find the answer yet. I have the following graph with a variable that ranges from 1 to 10. However I would like to change the y-axis to just show for example from 6 to 10. How can I do that?

Options like

graph bar (mean) score, over(country, label(labsize(vsmall))) over(date) yscale(range(6 10)) ylabel(6(1)10)

aren't working.

Cheers!

[Andrew Musau]

There is a way which I will not mention because it is a bad idea. See #2 https://www.statalist.org/forums/for...-for-bar-chart which recommends graph dot instead.

help graph dot

[Jean Jacques]

Thanks for the answer. So, first, I don't get what's the harm in a case like the one that I'm describing, in truncating or changing the axis. Second, using graph dot, still I can produce a similar output. This is my new code.

graph dot (mean) score, over(guest_country_code, label(labsize(vsmall))) over(match_date) yscale(range(6 10)) ylabel(6(1)10) vertical

which leads to:

Thanks !!

[Andrew Musau]

So, first, I don't get what's the harm in a case like the one that I'm describing, in truncating or changing the axis.

One issue is that starting a bar chart at some arbitrary positive non-zero value can lead to misleading visual interpretations. Note that a bar represents some quantity, either "how much" (such as a numerical measurement) or "how many" (such as a count or frequency).

second, using graph dot, still I can produce a similar output.

Please provide a data example using the dataex command.

[Jean Jacques]

Thanks for the answer. Agree with the first point. Indeed you can see many misleading graphs with different axis. Although in some cases (as the one I showed above) a simple bar chart is non-informative at all (and in this case are not quantities, are means).

About the second, here you have. Thanks

* Example generated by -dataex-. To install: ssc install dataex
clear
input double score byte guest_country_code long match_date
 7 0 0
10 1 0
 9 1 0
10 0 0
 7 1 0
10 1 0
10 1 0
 6 1 0
 5 0 0
10 1 0
 8 1 0
 6 1 0
 8 1 0
 6 1 0
 5 0 0
 8 1 0
10 0 0
 9 1 0
10 0 0
 9 1 0
 7 0 0
10 1 0
 5 0 1
10 0 1
 8 1 1
10 1 1
10 1 2
 7 0 2
 7 1 2
 7 0 2
end
label values guest_country_code country_label
label def country_label 0 "IT", modify
label def country_label 1 "UK", modify

[Nick Cox]

graph dot with the exclude0 option is ideal for your circumstances.

[Jean Jacques]

Great thanks. One more question. How can I make that the dots display a different color accoridig to whether they belong to a different guest_country_code (which is a dummy taking 0 or 1).

Up to now I do this:

graph dot (mean) score, over(guest_country_code, label(labsize(vsmall))) over(match_date) yscale(range(6 10)) ylabel(6(1)10) vertical exclude0  dots(msymbol(p))  marker(1, mcolor(green)) marker(2, mcolor(red))

without success. Thanks

[Nick Cox]

Option asyvars should help.

[Jean Jacques]

Thanks Nick, it worked. How should I proceed if I want to add confidence intervals? dot graph doesn't seems to accept it. Should I move to a two-way scatter? Would be possible to do that keeping the original chart as the one I have below:

graph dot (mean) score, over(guest_country_code, label(labsize(vsmall))) over(match_date, label(angle(45) labsize(small)) relabel(1 "- 5 days" 2 "- 4 days" 3 "- 3 days" 4 "- 2 days" 5 "- 1 day" 6 "Game day" 7 "+ 1 day" 8 "+ 2 day" 9 "+ 3 days" 10 "+ 4 days" 11 "+ 5 days" 12 "+ 6 days")) yscale(range(6 10)) ylabel(6(1)10) vertical exclude0  dots(msymbol(p)) asyvars marker(1, mcolor(gs9) msymbol(S)) marker(2, mcolor(gs3) msymbol(O)) ytitle("Average score")

Thanks
Jean

[Nick Cox]

See https://www.statalist.org/forums/for...ith-error-bars with a recent thread with a focus on error bars.

[Jean Jacques]

Sorry Nick, but I can still figure it out and I've been spending a good couple of hours on this.

I have computed the intervals

gen lb = .
gen ub = .

foreach c in "0" "1"{
foreach x in "1" "2" "3" "4" "5" "6" "7" "8" "9" "10" "11" {
ci means score if (match_date == `x') & (guest_country_code == `c')
replace lb = `r(lb)' if (match_date == `x') & (guest_country_code == `c')
replace ub = `r(ub)' if (match_date == `x') & (guest_country_code == `c')
}
}

but then I can replicate something similar to what you did in the other thread.

Here you have a snippet of my dataset.

* Example generated by -dataex-. To install: ssc install dataex
clear
input double score byte guest_country_code long match_date float(lb ub)
 7 1 1 7.996719 8.167284
10 1 1 7.996719 8.167284
 8 1 1 7.996719 8.167284
10 1 1 7.996719 8.167284
 7 1 1 7.996719 8.167284
 5 0 1  8.22762 8.408016
10 1 1 7.996719 8.167284
 8 0 1  8.22762 8.408016
10 0 1  8.22762 8.408016
10 0 1  8.22762 8.408016
 5 1 1 7.996719 8.167284
 6 1 1 7.996719 8.167284
 8 0 1  8.22762 8.408016
 8 1 1 7.996719 8.167284
10 0 1  8.22762 8.408016
 8 1 1 7.996719 8.167284
 9 1 1 7.996719 8.167284
 8 1 1 7.996719 8.167284
 8 0 1  8.22762 8.408016
 7 1 1 7.996719 8.167284
10 0 1  8.22762 8.408016
10 0 1  8.22762 8.408016
 9 0 1  8.22762 8.408016
 9 1 1 7.996719 8.167284
 5 1 1 7.996719 8.167284
 8 0 1  8.22762 8.408016
10 1 1 7.996719 8.167284
 3 1 1 7.996719 8.167284
10 0 1  8.22762 8.408016
 7 1 1 7.996719 8.167284
10 1 1 7.996719 8.167284
 6 0 1  8.22762 8.408016
 7 1 1 7.996719 8.167284
 9 1 1 7.996719 8.167284
 8 1 1 7.996719 8.167284
10 1 1 7.996719 8.167284
 8 1 1 7.996719 8.167284
10 1 1 7.996719 8.167284
 8 1 1 7.996719 8.167284
 6 0 1  8.22762 8.408016
 9 0 1  8.22762 8.408016
 8 1 1 7.996719 8.167284
 7 0 1  8.22762 8.408016
 8 0 1  8.22762 8.408016
10 1 1 7.996719 8.167284
 8 0 1  8.22762 8.408016
 5 1 1 7.996719 8.167284
 9 1 1 7.996719 8.167284
10 1 1 7.996719 8.167284
10 1 1 7.996719 8.167284
 8 0 1  8.22762 8.408016
10 1 1 7.996719 8.167284
10 1 1 7.996719 8.167284
10 1 1 7.996719 8.167284
 7 0 1  8.22762 8.408016
10 0 1  8.22762 8.408016
10 1 1 7.996719 8.167284
10 0 1  8.22762 8.408016
 8 0 1  8.22762 8.408016
10 0 1  8.22762 8.408016
 8 0 1  8.22762 8.408016
 5 1 1 7.996719 8.167284
 8 1 1 7.996719 8.167284
 9 1 1 7.996719 8.167284
10 1 1 7.996719 8.167284
10 0 1  8.22762 8.408016
 6 1 1 7.996719 8.167284
 6 0 1  8.22762 8.408016
10 1 1 7.996719 8.167284
 7 1 1 7.996719 8.167284
end
label values guest_country_code country_label
label def country_label 0 "IT", modify
label def country_label 1 "UK", modify

Can you help me?

[Nick Cox]

The graph from this looks fairly silly, but the hope is that the code will work fairly well for the complete dataset.

* Example generated by -dataex-. To install: ssc install dataex
clear
input double score byte guest_country_code long match_date float(lb ub)
 7 1 1 7.996719 8.167284
10 1 1 7.996719 8.167284
 8 1 1 7.996719 8.167284
10 1 1 7.996719 8.167284
 7 1 1 7.996719 8.167284
 5 0 1  8.22762 8.408016
10 1 1 7.996719 8.167284
 8 0 1  8.22762 8.408016
10 0 1  8.22762 8.408016
10 0 1  8.22762 8.408016
 5 1 1 7.996719 8.167284
 6 1 1 7.996719 8.167284
 8 0 1  8.22762 8.408016
 8 1 1 7.996719 8.167284
10 0 1  8.22762 8.408016
 8 1 1 7.996719 8.167284
 9 1 1 7.996719 8.167284
 8 1 1 7.996719 8.167284
 8 0 1  8.22762 8.408016
 7 1 1 7.996719 8.167284
10 0 1  8.22762 8.408016
10 0 1  8.22762 8.408016
 9 0 1  8.22762 8.408016
 9 1 1 7.996719 8.167284
 5 1 1 7.996719 8.167284
 8 0 1  8.22762 8.408016
10 1 1 7.996719 8.167284
 3 1 1 7.996719 8.167284
10 0 1  8.22762 8.408016
 7 1 1 7.996719 8.167284
10 1 1 7.996719 8.167284
 6 0 1  8.22762 8.408016
 7 1 1 7.996719 8.167284
 9 1 1 7.996719 8.167284
 8 1 1 7.996719 8.167284
10 1 1 7.996719 8.167284
 8 1 1 7.996719 8.167284
10 1 1 7.996719 8.167284
 8 1 1 7.996719 8.167284
 6 0 1  8.22762 8.408016
 9 0 1  8.22762 8.408016
 8 1 1 7.996719 8.167284
 7 0 1  8.22762 8.408016
 8 0 1  8.22762 8.408016
10 1 1 7.996719 8.167284
 8 0 1  8.22762 8.408016
 5 1 1 7.996719 8.167284
 9 1 1 7.996719 8.167284
10 1 1 7.996719 8.167284
10 1 1 7.996719 8.167284
 8 0 1  8.22762 8.408016
10 1 1 7.996719 8.167284
10 1 1 7.996719 8.167284
10 1 1 7.996719 8.167284
 7 0 1  8.22762 8.408016
10 0 1  8.22762 8.408016
10 1 1 7.996719 8.167284
10 0 1  8.22762 8.408016
 8 0 1  8.22762 8.408016
10 0 1  8.22762 8.408016
 8 0 1  8.22762 8.408016
 5 1 1 7.996719 8.167284
 8 1 1 7.996719 8.167284
 9 1 1 7.996719 8.167284
10 1 1 7.996719 8.167284
10 0 1  8.22762 8.408016
 6 1 1 7.996719 8.167284
 6 0 1  8.22762 8.408016
10 1 1 7.996719 8.167284
 7 1 1 7.996719 8.167284
end
label values guest_country_code country_label
label def country_label 0 "IT", modify
label def country_label 1 "UK", modify

preserve 

statsby, by(guest_country_code match_date) clear: ci means score  

* delta is at choice 
local delta = 0.2 
gen where = match_date + cond(guest_country_code == 1, `delta', -`delta')

su match_date, meanonly 
local dmin = r(min)
local dmax = r(max)

forval x = `dmin'/`dmax' { 
    local left = `x' - `delta'
    local right = `x' + `delta'
    local xmla `xmla' `left' "IT" `right' "UK"
}

scatter mean where, xla(`dmin'/`dmax') || rcap ub lb where, xmla(`xmla', tlc(none) labsize(medsmall)) legend(off) ///
ytitle(mean score and 95% confidence interval) xtitle(match date)

[Jean Jacques]

Great, thanks.

Just in case someone else wants to know the code to leave it is what I was looking for, here you have the code:

preserve 

statsby, by(guest_country_code match_date) clear: ci means score  

* delta is at choice 
local delta = 0.2 
gen where = match_date + cond(guest_country_code == 1, `delta', -`delta')



su match_date, meanonly 
local dmin = r(min)
local dmax = r(max)

forval x = `dmin'/`dmax' { 
    local left = `x' - `delta'
    local right = `x' + `delta'
    local xmla `xmla' `left' "IT" `right' "UK"
}


scatter mean where, xmla(`dmin'/`dmax', labsize(vsmall) angle(90) tlc(none) tlength(*9)) mcolor(gs2) msymbol(s) xmlabel(1 "- 5 days" 2 "- 4 days" 3 "- 3 days" 4 "- 2 days" 5 "- 1 day" 6 "Game day" 7 "+ 1 day" 8 "+ 2 day" 9 "+ 3 days" 10 "+ 4 days" 11 "+ 5 days" 12 "+ 6 days") || rcap ub lb where, lcolor(gs2) xla(`xmla', grid glcolor(gs7) glpattern(dot) labsize(small)) graphregion(color(white)) bgcolor(white) ylab(7(1)10, nogrid) xtitle("") legend(off)