Scatter plot with time line

[Adriano Amaral]

Hi there,

I would like to create a scatter plot between two variables, for example income vs consumption, but I want to adjust the result over a quarterly timeline. I could not find anywhere an answer for this specific problem.

Regards, Adriano

[Nick Cox]

I understand a scatter plot with points and a line plot with time on one axis. I don't follow quite what you're seeking. Can you give an example, a reference, or somehow more detail?

[Adriano Amaral]

A generated a scatter plot for 2 variables, consumption and income, and segmented it for before and after an event. The event occurs at the 3rd quarter - I have a 75% before and a 25% after in terms of time. However, the scatter plot locates the breakpoint on the middle of the x axis. I would like to adjust the spread to the real timeline.

[Adriano Amaral]

I used the following command

twoway (scatter housecons govtotaldebt ) (lfit housecons govtotaldebt if quarter <60, sort) (lfit housecons govtotaldebt if quarter >=60, sort), xline(60)

[Nick Cox]

xline(60) presumably makes no sense to Stata as govtotaldebt is plotted on the x axis in all your twoway calls. Perhaps you should use the value of that variable for quarter 60.

[Nick Cox]

More thoughts:

0. It seems that you are not using Stata 19. Please tell us the version you are using. See FAQ Advice. https://www.statalist.org/forums/help#version

1. Consider joining data points in time order within each group.
2. Consider open marker symbols Oh which tolerate overlap better than default O.
3. Consider adding time labels for first and last quarters and quarters at the end and beginning of each group.
4. Divide consumption and debt by 1 million and state units on the graph.

Sample code follows. Marker label positions may need tweaking.

* you will not need to be told how to divide by 1 million 

su govtotaldebt if inlist(quarter, 59, 60), meanonly 
local where = r(mean) 

su quarter, meanonly 
local first = r(min)
local last = r(max) 

* what follows is all one command 
twoway 
(scatter housecons govtotaldebt, ms(Oh) ytitle(Income (units here)) xtitle(Consumption (units here)))  
(lfit housecons govtotaldebt if quarter <60, sort) 
(lfit housecons govtotaldebt if quarter >=60, sort), xline(`where')
(scatter housecons govtotaldebt if quarter == `first', ms(none) mla(quarter) mlabpos(9)) 
(scatter housecons govtotaldebt if quarter == 59, ms(none) mla(quarter) mlabpos(3)) 
(scatter housecons govtotaldebt if quarter == 60, ms(none) mla(quarter) mlabpos(9)) 
(scatter housecons govtotaldebt if quarter == `last', ms(none) mla(quarter) mlabpos(9))

[Adriano Amaral]

xline(60) presumably makes no sense to Stata as govtotaldebt is plotted on the x axis in all your twoway calls. Perhaps you should use the value of that variable for quarter 60.

I tried that first, but didn't work. The result was the same.

[Adriano Amaral]

I have Stata 12. Thank you so much for taking the time to respond. I will explore the options suggested.

[Nick Cox]

#7 is not a problem that I can decode ("didn't work").

I use the || notation for combining twoway calls. There are in my view quite enough parentheses to keep track of. There are typos in #6: this should be better as the final command. .

twoway
scatter housecons govtotaldebt, ms(Oh) ytitle(Income (units here)) xtitle(Consumption (units here))
|| lfit housecons govtotaldebt if quarter <60
|| lfit housecons govtotaldebt if quarter >=60, xline(`where')
|| scatter housecons govtotaldebt if quarter == `first', ms(none) mla(quarter) mlabpos(9)
|| scatter housecons govtotaldebt if quarter == 59, ms(none) mla(quarter) mlabpos(3)
|| scatter housecons govtotaldebt if quarter == 60, ms(none) mla(quarter) mlabpos(9)
|| scatter housecons govtotaldebt if quarter == `last', ms(none) mla(quarter) mlabpos(9)

Version 12 may not be a problem, but you need to try. I've lost access to Stata 12 long since.

By the way, I am guessing that your quarter variable is just 1 up for your data, but using Stata quarterly dates may help the reader.