Graph counts of categorical variable by year

[Varsha Vaishnav]

Hello,

I have the counts of a categorical variable by year. Please suggest a good way to graph such data. An example:

* Example generated by -dataex-. For more info, type help dataex
clear
input byte finyear str39 workcategoryname int _freq
10 "Anganwadi/Other Rural Infrastructure"    2
11 "Anganwadi/Other Rural Infrastructure"   14
12 "Anganwadi/Other Rural Infrastructure"    7
13 "Anganwadi/Other Rural Infrastructure"    6
14 "Anganwadi/Other Rural Infrastructure"   22
15 "Anganwadi/Other Rural Infrastructure"    5
16 "Anganwadi/Other Rural Infrastructure"   60
17 "Anganwadi/Other Rural Infrastructure"  105
18 "Anganwadi/Other Rural Infrastructure"  152
19 "Anganwadi/Other Rural Infrastructure"  190
20 "Anganwadi/Other Rural Infrastructure" 2518
21 "Anganwadi/Other Rural Infrastructure"  764
22 "Anganwadi/Other Rural Infrastructure"  505
23 "Anganwadi/Other Rural Infrastructure"  213
24 "Anganwadi/Other Rural Infrastructure"  526
25 "Anganwadi/Other Rural Infrastructure"  112
10 "Bharat Nirman Sewa Kendra"              86
11 "Bharat Nirman Sewa Kendra"             271
12 "Bharat Nirman Sewa Kendra"             345
13 "Bharat Nirman Sewa Kendra"             162
14 "Bharat Nirman Sewa Kendra"             235
15 "Bharat Nirman Sewa Kendra"              31
16 "Bharat Nirman Sewa Kendra"              51
17 "Bharat Nirman Sewa Kendra"              22
18 "Bharat Nirman Sewa Kendra"              19
19 "Bharat Nirman Sewa Kendra"               2
13 "Coastal Areas"                          13
14 "Coastal Areas"                           4
16 "Coastal Areas"                           2
17 "Coastal Areas"                           3
19 "Coastal Areas"                           4
20 "Coastal Areas"                           2
22 "Coastal Areas"                           1
24 "Coastal Areas"                           1
10 "Drought Proofing"                      406
11 "Drought Proofing"                      478
12 "Drought Proofing"                      153
13 "Drought Proofing"                      163
14 "Drought Proofing"                      420
15 "Drought Proofing"                      299
16 "Drought Proofing"                      463
17 "Drought Proofing"                      867
18 "Drought Proofing"                     1062
19 "Drought Proofing"                      585
20 "Drought Proofing"                     4292
21 "Drought Proofing"                     2040
22 "Drought Proofing"                     2526
23 "Drought Proofing"                     1405
24 "Drought Proofing"                     5624
25 "Drought Proofing"                      437
end

Thank you.

[Clyde Schechter]

Without know just what questions you hope to answer with this data, I cannot suggest a "best" way of doing this. But here is one way that puts all of the data in a single graph, readably, and enables you to see the variation in each category.

encode workcategoryname, gen(category)

xtset category finyear

xtline _freq, yscale(log) ylabel(1 10 100 1000 10000) overlay

The log scale on the vertical axis was chosen because with an ordinary linear scale, the small values for Coastal Areas and Bharat Nirman Sewa Kendra get squashed to the point that their variation over time cannot be seen.

But depending on your goals, there are many other ways to display this data that might be more suitable.

[Nick Cox]

Clyde Schechter is as usual entirely correct, especially that we need more information to give good advice.

An easy guess is that you have about 16 different years. It's harder to guess how many categories you have. It's all too easy to get to the point where a large fraction of a display is taken up by a legend.

[Varsha Vaishnav]

I have 16 categories.

[Hemanshu Kumar]

With as many as 16 categories, graphs made in the style of #2 might get a tad messy. Perhaps try sparkline or multiline, both contributed by Nick Cox and available via ssc install?

With your example data,

sparkline _freq finyear, over(workcategoryname)

produces

Unlike sparkline, multiline does not have an over() option, so we need to first create the requisite variables, and then graph them:

separate _freq, by(workcategoryname) veryshortlabel
multiline _freq? finyear, missing

to get

[Varsha Vaishnav]

Thank you, Hemanshu.