Stacked Bar Chart: Date Labels dont work

[Felix Meier]

Hello everyone,

I created the stacked bar chart attached using the variables and the code below:

graph bar, over(party) over(quarter_created, label(angle(45))) missing stack asyvars

My problem is that the variable on the x-axis doesn't take the labels I assigned to it, but number instead. It is in float format and has date-labels for each quarter, which I assigned using this code:

gen date_created = date(created_at, "YMD hms")
format date_created %td
gen quarter_created = qofd(date_created)
format quarter_created %tqCCYY-!Qq

Where created_at originally is in string format. Example: "2008-05-15 11:26:35"
And quarter_created is in float format with labels like: "2008q3"

I would be very happy if you could help me with my problem.

[Nick Cox]

Wrong description. Your date variable doesn't have value labels attached at all, but a date display format, which graph bar won't honour. That's a bug, or at least a limitation. You could assign each date its own value label, but that still leaves problems.

You would be immensely better off with twoway bar instead when you could just use fewer axis labels. Indeed the categorisation does not work well either, but that is another story.

Give some example data, and I will illustrate.

[Felix Meier]

Thank you so much for your answer! I have been trying to use twoway bar on my own but didn't manage to get the results above with the legend and colors. Here is some random sample data from my project (thereby it's not really important if the x-axis is in percent of all observations or just the total number of observations per quarter created. To make it easier I also included the quarter variable without the labels.

quarter_created    party_id    quarter_created_nolabels
2011-Q4    partyC    207
2009-Q2    partyC    197
2018-Q2    partyE    233
2009-Q2    partyF    197
2010-Q2    partyF    201
2009-Q2    partyE    197
2013-Q2    partyH    213
2017-Q2    partyG    229
2018-Q2    partyB    233
2016-Q4    partyB    227
2017-Q4    partyB    231
2017-Q3    partyH    230
2014-Q2    partyB    217
2016-Q2    partyB    225
2011-Q3    partyC    206
2017-Q1    partyH    228
2008-Q3    partyF    194
2015-Q3    partyH    222
2009-Q2    partyF    197
2009-Q2    partyE    197
2013-Q2    partyG    213
2017-Q2    partyE    229
2012-Q4    partyH    211
2018-Q1    partyE    232
2009-Q4    partyG    199
2009-Q1    partyG    196
2008-Q4    partyH    195
2014-Q2    partyC    217
2013-Q1    partyE    212
2009-Q4    partyB    199
2008-Q4    partyF    195
2008-Q3    partyH    194
2009-Q4    partyC    199
2016-Q4    partyB    227
2014-Q4    partyH    219
2011-Q2    partyH    205
2009-Q2    partyC    197
2009-Q2    partyE    197
2010-Q2    partyH    201
2009-Q4    partyE    199
2017-Q2    partyB    229
2014-Q3    partyH    218
2015-Q1    partyB    220
2018-Q1    partyE    232
2017-Q1    partyH    228
2009-Q2    partyH    197
2016-Q4    partyC    227
2018-Q1    partyE    232
2018-Q1    partyF    232
2012-Q3    partyC    210

Thank you very much!

[Nick Cox]

Thanks for the data example, but please use dataex (FAQ Advice #12)

I use tabplot (Stata Journal) which must be installed first.

https://www.statalist.org/forums/for...updated-on-ssc is perhaps the best way in if this command is new to you.

Here's my code, most of which is just creating your data example.

clear 
input str7 (quarter_created    party_id)    quarter_created_nolabels
2011-Q4    partyC    207
2009-Q2    partyC    197
2018-Q2    partyE    233
2009-Q2    partyF    197
2010-Q2    partyF    201
2009-Q2    partyE    197
2013-Q2    partyH    213
2017-Q2    partyG    229
2018-Q2    partyB    233
2016-Q4    partyB    227
2017-Q4    partyB    231
2017-Q3    partyH    230
2014-Q2    partyB    217
2016-Q2    partyB    225
2011-Q3    partyC    206
2017-Q1    partyH    228
2008-Q3    partyF    194
2015-Q3    partyH    222
2009-Q2    partyF    197
2009-Q2    partyE    197
2013-Q2    partyG    213
2017-Q2    partyE    229
2012-Q4    partyH    211
2018-Q1    partyE    232
2009-Q4    partyG    199
2009-Q1    partyG    196
2008-Q4    partyH    195
2014-Q2    partyC    217
2013-Q1    partyE    212
2009-Q4    partyB    199
2008-Q4    partyF    195
2008-Q3    partyH    194
2009-Q4    partyC    199
2016-Q4    partyB    227
2014-Q4    partyH    219
2011-Q2    partyH    205
2009-Q2    partyC    197
2009-Q2    partyE    197
2010-Q2    partyH    201
2009-Q4    partyE    199
2017-Q2    partyB    229
2014-Q3    partyH    218
2015-Q1    partyB    220
2018-Q1    partyE    232
2017-Q1    partyH    228
2009-Q2    partyH    197
2016-Q4    partyC    227
2018-Q1    partyE    232
2018-Q1    partyF    232
2012-Q3    partyC    210
end 

replace party_id = substr(party_id, -1, 1)
set scheme s1color 
tabplot party_id quarter_created_nolabels, xasis xla(, noticks ang(v) format("%tqCY!Qq")) xtitle("") bfcolor(none) showval ytitle("")

I suggest that this is a more direct display which avoids a confusing multi-colour stacking with the need to decode using a legend or key and much mental back and forth. I have no idea what the data show: perhaps the full dataset is enormously larger and there is some reason why you use coy labels B C D E F G H.

[Felix Meier]

Thank you again! The graph you suggest is interesting, though I still believe that a stacked bar chart would be better. This is the case since the main information of the graph is the height of the bars (in my data it means when politicians entered twitter and whether this happened just before the elections). Which party they belong to is just a sidenote and interesting, but not super important. For this reason it'd be great to have a stacked bar chart with coloring by party and height related to how many people entered in which quarter. Is there any way to do this in Stata? The total dataset is not super large and only consists of 460 observations.

[Nick Cox]

Here's something to get you started:

replace party_id = substr(party_id, -1, 1)
drop quarter_created 
rename quarter quarter 
format quarter %tq 
contract party_id quarter, zero  

separate _freq, by(party_id) 
mvencode _freq*, mv(0) override 

gen bar1 = _freq1

forval j = 2/6 {
    local J = `j' - 1 
    gen bar`j' = bar`J' + _freq`j' 
} 

twoway bar bar1 quarter || /// 
rbar bar1 bar2 quarter  || /// 
rbar bar2 bar3 quarter  || /// 
rbar bar3 bar4 quarter  || /// 
rbar bar4 bar5 quarter  || /// 
rbar bar5 bar6 quarter  || /// 
, legend(order(1 "B" 2 "C" 3 "E" 5 "F" 6 "G"))