Plotting Proportions of Groups Over Time

[Jared Greathouse]

Consider the following

* Example generated by -dataex-. For more info, type help dataex
clear
input int year float(over50 grassnew)
1973 1 0
1973 1 0
1973 0 1
1973 0 0
1973 1 0
1973 0 0
1973 0 0
1973 0 1
1973 1 0
1973 0 0
1973 1 0
1973 0 0
1973 1 0
1973 0 0
1973 0 0
1973 1 1
1973 0 1
1973 0 0
1973 1 1
1973 1 0
1973 1 0
1973 0 0
1973 0 0
1973 1 0
1973 1 1
1973 0 0
1973 1 0
1973 0 0
1973 1 0
1973 0 1
1973 1 0
1973 1 0
1973 1 0
1973 1 0
1973 0 1
1973 0 0
1973 0 0
1973 0 0
1973 0 0
1973 1 0
1973 0 1
1973 1 0
1973 1 0
1973 0 1
1973 0 0
1973 0 1
1973 1 1
1973 0 0
1973 0 1
1973 0 0
1973 0 0
1973 0 1
1973 1 0
1973 0 0
1973 1 0
1973 0 0
1973 1 0
1973 0 0
1973 1 0
1973 1 0
1973 0 1
1973 0 0
1973 1 0
1973 1 0
1973 1 0
1973 0 0
1973 0 0
1973 0 0
1973 0 0
1973 0 0
1973 0 0
1973 0 0
1973 0 1
1973 0 1
1973 0 1
1973 1 0
1973 0 0
1973 0 0
1973 0 1
1973 1 0
1973 1 1
1973 0 0
1973 1 0
1973 1 0
1973 1 0
1973 1 0
1973 0 1
1973 0 1
1973 0 0
1973 0 0
1973 0 0
1973 0 1
1973 1 0
1973 1 0
1973 0 0
1973 0 0
1973 0 0
1973 1 0
1973 0 0
1973 1 0
end

I have survey respondent data, from 1973 to 2018. I want to plot, by age group (over50, over/under 50, 1, 0 respectively) how many people support legalizing cannabis (grassnew, 1 who supports, else 0). My first thought is to collapse this, so we get the proportion/mean of people who say yes in each year, and in each age group... But this would be tricky, since I'd need to create more variables/frames to do this.

is there a more elaborate way to do this, simply to plot the group level support trends over time?

[Nick Cox]

egen toshow = mean(grassnew), by(over50 year)

gets you the proportions and using instead

mean(100 * grassnew)

gets you the percents.

Now that's only 46 X 2 distinct data points so to avoid massive overplotting

egen tag = tag(over50 year) 


line toshow year if over50 & tag || line toshow year if !over50 & tag , legend(order(1 "over 50" 2 "under 50"))

Which way do 50 year olds go?

[Clyde Schechter]

So, the first way that comes to my mind for this is

collapse (mean) grassnew, by(year over50)
reshape wide grassnew, i(year) j(over50)
graph twoway connect grassnew* year

And I take it you saw this, too. I guess I don't grasp what you consider tricky about this. If you object to the reshape, you could do this:

collapse (mean) grassnew, by(year over50)
twoway (connect grassnew year if over50 == 0) (connect grassnew year if over50 == 1)

[Leonardo Guizzetti]

Do you have survey weights or design elements to consider?

[Jared Greathouse]

No thankfully🤣 either way, Nick and Clyde's ways both worked!