Applying different colors to data points when using aaplot

[Stephen Okiya]

Hi Stata Users,
I am using a Stata user written command aaplot and would like to colour code the points using the category variable. I am using the syntax and the dataset below

* Example generated by -dataex-. For more info, type help dataex
clear
input byte(value_2014 growth category)
21  1 1
11  5 1
20  6 2
16  7 3
14  3 3
18  3 1
19  1 2
20  4 3
21 10 1
22 11 2
10  2 3
17  5 3
18  9 3
12  7 3
11  8 2
10  6 2
14  5 2
17  7 1
20 11 1
22 12 1
21 15 1
15  2 1
15  5 2
17  8 2
19  9 2
10  6 3
10  4 3
14  4 3
20  5 1
16  7 2
19  8 3
18  9 1
10  3 1
11  9 1
12  8 2
14  7 2
15  3 2
18  5 3
19  5 1
13  3 2
14  9 3
15  2 3
15  1 3
16 10 2
17 12 1
19 11 2
18 16 3
12  6 2
10  9 2
20  2 2
end

aaplot growth value_2014

Thanks in advance!

[Nick Cox]

aaplot is from SSC. You could do something like:

separate growth, by(category) veryshortlabel 
aaplot growth value_2014, ms(none) backdrop(scatter growth? value_2014, ms(Oh + Th)) xla(10(2)22) legend(on order(1 2 3) row(1) ring(0) pos(11))

There's plenty of scope to improve on that: define variable labels, round results to fewer decimal places, and so on.

[Stephen Okiya]

Thanks Nick Cox for the incredible help. This is really helpful.

[Stephen Okiya]

I am trying to have all points to be circles using the syntax below but the results aren't as intended

separate growth, by(category) veryshortlabel 
aaplot growth value_2014, ms(none) backdrop(scatter growth? value_2014, ms(Oh + Oh + Oh)) xla(10(2)22) legend(on order(1 2 3) row(1) ring(0) pos(11))

Any help will be greatly appreciated.

[Nick Cox]

You're asking for open circie, plus, open circle, plus, open circle. So, don't ask for plus. In this context the + does not mean "and". It means that you want the marker symbol to be a plus sign +.

ms(Oh Oh Oh)

or

ms(Oh ..)

[Stephen Okiya]

Thanks so much Nick Cox for the guidance. I sincerely appreciate your help.

[Nick Cox]

Your data in #1 show many identical data points. In these circumstances circles even with different colours can't work well unless you make them have different sizes or add jitter.

[Stephen Okiya]

Thank you so much for the advice - it makes sense to use different symbols to address overlap of points

[Stephen Okiya]

I have encountered an outlier that I would like to exclude in the analysis. Below is a modified version of data to reflect the case

* Example generated by -dataex-. For more info, type help dataex
clear
input byte(value_2014 growth category)
21  1 1
11  5 1
20  6 2
16  7 3
14  3 3
18  3 1
19  1 2
20  4 3
21 10 1
22 11 2
10  2 3
17  5 3
18  9 3
12  7 3
11  8 2
10  6 2
14  5 2
17  7 1
20 11 1
22 12 1
21 15 1
15  2 1
15  50 2
17  8 2
19  9 2
10  6 3
10  4 3
14  4 3
20  5 1
16  7 2
19  68 3
18  9 1
10  3 1
11  9 1
12  8 2
14  7 2
15  3 2
18  5 3
19  5 1
13  3 2
14  9 3
15  2 3
15  1 3
16 10 2
17 12 1
19 11 2
18 16 3
12  6 2
10  9 2
20  2 2
end

I am using the following code but it seems not to work

 separate growth, by(category) veryshortlabel  aaplot growth value_2014 if growth < 20, ms(none) backdrop(scatter growth? value_2014, ms(Oh + Th)) xla(10(2)22) legend(on order(1 2 3) row(1) ring(0) pos(11))

Thanks in advance

[Nick Cox]

You need to apply the if qualifier within the addplot() call too.