bar with confidence interval?

[River Huang]

Dear All, I find this question here (http://bbs.pinggu.org/thread-6292609-1-1.html). How to draw a graph like this. (Unfortunately, no data set is available.)

[Nick Cox]

twoway bar ... || rcap ...

Better to show more data on the plot (and if values should always be positive calculate confidence intervals on a more appropriate scale).

Search phrases: dynamite plots, detonator plots, plunger plots

http://biostat.mc.vanderbilt.edu/wik...de/Poster3.pdf is worth reading, even though the author's main target is a step beyond what you're showing.

[River Huang]

Dear Nick, I guess I need more assistance. Suppose that I have the data as (webuse grunfeld, clear)

* Example generated by -dataex-. To install: ssc install dataex
clear
input float(company year invest)
1 1935  317.6
1 1936  391.8
1 1937  410.6
1 1938  257.7
1 1939  330.8
1 1940  461.2
1 1941    512
1 1942    448
1 1943  499.6
1 1944  547.5
1 1945  561.2
1 1946  688.1
1 1947  568.9
1 1948  529.2
1 1949  555.1
1 1950  642.9
1 1951  755.9
1 1952  891.2
1 1953 1304.4
1 1954 1486.7
2 1935  209.9
2 1936  355.3
2 1937  469.9
2 1938  262.3
2 1939  230.4
2 1940  361.6
2 1941  472.8
2 1942  445.6
2 1943  361.6
2 1944  288.2
2 1945  258.7
2 1946  420.3
2 1947  420.5
2 1948  494.5
2 1949  405.1
2 1950  418.8
2 1951  588.2
2 1952  645.5
2 1953    641
2 1954  459.3
3 1935   33.1
3 1936     45
3 1937   77.2
3 1938   44.6
3 1939   48.1
3 1940   74.4
3 1941    113
3 1942   91.9
3 1943   61.3
3 1944   56.8
3 1945   93.6
3 1946  159.9
3 1947  147.2
3 1948  146.3
3 1949   98.3
3 1950   93.5
3 1951  135.2
3 1952  157.3
3 1953  179.5
3 1954  189.6
4 1935  40.29
4 1936  72.76
4 1937  66.26
4 1938   51.6
4 1939  52.41
4 1940  69.41
4 1941  68.35
4 1942   46.8
4 1943   47.4
4 1944  59.57
4 1945  88.78
4 1946  74.12
4 1947  62.68
4 1948  89.36
4 1949  78.98
4 1950 100.66
4 1951 160.62
4 1952    145
4 1953 174.93
4 1954 172.49
5 1935  39.68
5 1936  50.73
5 1937  74.24
5 1938  53.51
5 1939  42.65
5 1940  46.48
5 1941   61.4
5 1942  39.67
5 1943  62.24
5 1944  52.32
5 1945  63.21
5 1946  59.37
5 1947  58.02
5 1948  70.34
5 1949  67.42
5 1950  55.74
5 1951   80.3
5 1952   85.4
5 1953   91.9
5 1954  81.43
end
format %ty year

How can I draw a bar graph for `invest' for each year like the one above? Thanks.

[Nick Cox]

You need to summarize the data first. Then as said start with twoway bar. Or consider first using statsby as in http://www.stata-journal.com/sjpdf.h...iclenum=gr0045 (but note that the syntax for ci has changed since that paper).

[Marcos Almeida]

I believe - serrbar - can present the situation "comme il faut", I mean, it can avoid the dynamite-plot's pitfalls.

You may wish to try this:

statsby mean=r(mean) sd=r(sd) size=r(N), by(year):  summarize invest
serrbar mean sd year

[River Huang]

Hi, Nick: Thank you. I will check it out.

[River Huang]

Hi Marcos, Thanks for this interesting suggestion.

[River Huang]

Stata code is

webuse grunfeld, clear
statsby mean=r(mean) ub=r(ub) lb=r(lb) N=r(N), by(year) clear: ci mean invest
twoway bar mean year, color(gs12) barw(0.5) || rcap ub lb year, legend(off)

A figure is as follow:

[Chris Larkin]

ssc install cibar provides a nice pre-packaged alternative

[River Huang]

Dear Chris, Many thanks for suggesting this nice command.