How to add a secondary y-axis to a bar chart with the graph bar command

[Giovanni Palmioli]

Hi everyone,

I am new here and would appreciate your help with the following bar chart.

The chart has 3 (year) bars per id (string variable). I am wondering if anyone knows how to add a secondary y-axis with the graph bar command.

Should the graph bar not allow for a secondary y-axis, would the twoway bar command be the solution? The id can be transformed to a numerical value but should eventually be relabelled with the full name on the x axis.

graph bar (mean) var, over(year) over(id)

Cheers,
Giovanni

[Nick Cox]

There is no scope for a secondary y axis with graph bar. I don't know what you want to do with it but twoway bar sounds a more likely direction.

Please read http://www.statalist.org/forums/help#stata for advice on (1) showing example data and code (2) showing graphs as .png attachments, not photos.

Judging from your example graph, log scale is the answer, not a secondary axis.

[Giovanni Palmioli]

Hi Nick,

I would like to follow up on this post. Thanks for the suggestion but I cannot use a log scale because the values are percentages and I need to keep them as they are.

I would like to have 3 stacked bars (year 2006, 2010, 2010) next to each other per id_s and use a secondary y-axis for some id_s (in particular 1 and 7). How would you use twoway bar(if it is a better option than graph bar)?

I am interested only in the pairs with id_h=100 but would like to keep all pairs. The panel below is a little different from the one I used for the chart in my original post.

Your inputs would be much appreciated.

 
id_s
id_h
year
exp_a_perc
exp_b_perc

1
100
2006
18.99203
17.12362

1
100
2010
31.92924
35.2694

1
100
2014
18.900452
19.852894

2
100
2006
3.85358
1.0512108

2
100
2010
8.024672
2.316856

2
100
2014
7.654442
1.6625088

3
100
2006
0.9600852
2.04596

3
100
2010
1.84154
2.861238

3
100
2014
0.997911
1.6934196

4
100
2006
0.0183314
0.517915

4
100
2010
8.952862
1.3665348

4
100
2014
5.784374
0.6271664

5
100
2006
3.041822
5.23905

5
100
2010
6.202086
9.874898

5
100
2014
1.5041632
2.863996

6
100
2006
0.5564346
0.1237718

6
100
2010
0.953272
0.4472336

6
100
2014
0.7111274
0.4214042

7
100
2006
16.805914
28.41752

7
100
2010
32.60148
35.03638

7
100
2014
8.408258
15.419746

8
100
2006
1.3177232
1.1193254

8
100
2010
2.686618
1.76735

8
100
2014
2.52161
1.423866

..
200
..
..
..

..
200
..
..
..

…
200
..
..
..

[Nick Cox]

Thanks for the data example. Looking at your data I would say that the problem is not so much accommodating the large values as showing the small values properly.

I tried this:

clear 

set scheme s1color 

input id_s    id_h    year    exp_a_perc    exp_b_perc
1    100    2006    18.99203    17.12362
1    100    2010    31.92924    35.2694
1    100    2014    18.900452    19.852894
2    100    2006    3.85358    1.0512108
2    100    2010    8.024672    2.316856
2    100    2014    7.654442    1.6625088
3    100    2006    0.9600852    2.04596
3    100    2010    1.84154    2.861238
3    100    2014    0.997911    1.6934196
4    100    2006    0.0183314    0.517915
4    100    2010    8.952862    1.3665348
4    100    2014    5.784374    0.6271664
5    100    2006    3.041822    5.23905
5    100    2010    6.202086    9.874898
5    100    2014    1.5041632    2.863996
6    100    2006    0.5564346    0.1237718
6    100    2010    0.953272    0.4472336
6    100    2014    0.7111274    0.4214042
7    100    2006    16.805914    28.41752
7    100    2010    32.60148    35.03638
7    100    2014    8.408258    15.419746
8    100    2006    1.3177232    1.1193254
8    100    2010    2.686618    1.76735
8    100    2014    2.52161    1.423866
end 

tabplot year id_s [iw=exp_a_perc], showval(format(%2.1f))

more 

tabplot year id_s [iw=exp_b_perc], showval(format(%2.1f))

more 

bysort year (exp_a_perc) : gen order = _N - _n + 1 if year == 2006
bysort id_s (year): replace order = order[1]
labmask order, values(id_s)

tabplot year order [iw=exp_a_perc], showval(format(%2.1f))

Here tabplot must be installed from SSC http://www.statalist.org/forums/foru...updated-on-ssc

Given several small values, presumably of real interest too, "looking them up" is eased by a hybrid graph and table.

Here I show your two variables, and with a guess that the identifier may be arbitrary, an ordering on values in 2006.

labmask should be installed from the Stata Journal files.

[Giovanni Palmioli]

Nick,

Many thanks for your prompt feedback.

Sorry, my previous post was not very clear maybe. I am looking to create a bar chart similar to the one I attached in my first post. Can you see the attachment?

Hope this clarifies my previous post.

[Nick Cox]

You were entirely clear in broad terms. You want one set of stacked bars related to one axis and another to another axis. I have to declare zero interest in writing code for that, as I think it's quite unnecessary.

Naturally, the thread is entirely open for other suggestions..