Pen's Parade/ Paintdrip Chart

[Nick Avis]

Hi everyone,

I am trying to create a Pen's parade or Paintdrip Chart in Stata for cumulative income minus a given household expenditure for each household. I have tried creating plotting cumulative frequency graphs for both variables with limited success and also the use of the 'alorenz' module. I have inserted a picture below of the kind of thing I would like to replicate with a different dataset. Any help is greatly appreciated as I'm a bit of a Stata novice.

Thanks, Nick

[Andrew Lover]

help twoway_dropline

should get you started.

[Nick Avis]

Thanks Andrew,

I've taken a look and it's the closest I've come to so far. The aim is to have a line falling from the ratio of income to poverty level pre-payment (POVLEV18) to the ratio of income to poverty level post-payment (NMPOVLEV18), and whether the line drops below 100 on the y-axis.

cumul POVLEV18 if !missing(POVLEV18), gen(POVcum)
twoway dropline NMPOVLEV18 POVcum

An issue is I have 10,000 observations so even reducing the width of the dropline does not improve clarity. I've attached a picture of the result.

[Tarun Choudhary]

You will need to have three variables for a) ordering the individuals based on pre-OOP household consumption b) pre-OOP household consumption c) post-OOP household consumption
The following code should then get you the desired results

sort pre_oop
gen order = _n
twoway line pre_oop order || line post_oop order,legend(pos(6)

You will need to tweak a few options to get the exact graph that you want

[Nick Avis]

You will need to have three variables for a) ordering the individuals based on pre-OOP household consumption b) pre-OOP household consumption c) post-OOP household consumption
The following code should then get you the desired results

sort pre_oop
gen order = _n
twoway line pre_oop order || line post_oop order,legend(pos(6)

You will need to tweak a few options to get the exact graph that you want

Hi Tarun,

Thank you for your suggestion. I've applied your advice and it looks great.

I've attached a picture, needs a little bit of fiddling but certainly what I was looking for!
Thanks again.

[Muhamad Farhan]

Hi Nick Avis, do you mind to share me how you get into this graph? Currently trying to generate a similar graph in Indonesia's case. Thank you in advance.

ps: I've tried using the above suggestion, but apparently there are several commands missing to get into your final graph.

[Nick Cox]

The advice in #4 seems sufficient to get the graph in #5 provided that one variable is always less than or equal to the variable you're ordering on. Otherwise here is some different technique in a self-contained script you can run.

clear 
set obs 50 
set seed 314159265

gen y1 = exp(rnormal(0,1))
gen y2 = y1 - exp(rnormal(0, 0.5))

sort y1
gen order = _n
twoway rspike y1 y2 order, name(G1, replace)
twoway rspike y1 y2 order || line y1 order , name(G2, replace)