I am trying to draw a common pre-trend graph for a diff-in-diff design. My treatment group is made of one only state, while the control group is made of three states. The data are a panel so that I have obseervations for each unit (state) over the same period of time. I'll post here a portion of it
Now what I would like to plot are just two lines, one for the treated group and one for the control group (three of those state in aggregate, e.g. NSW, NT and Qld). What I need to plot is the ratio "cases/pop".
My code by now is
where treated variable is equal to 1 for state ACT and equal to 0 for state NSW, NT and Qld. But the result is a plot with 4 lines, one for each state instead of one for the treated group and one for the control group in aggregate
What is the more meaningful way of doing it? Should I manually aggregate the data (i.e. summing cases and pop by year and create a new "control" state) or is there a easier way of doing it?
Code:
state year cases pop ACT 1971 35 151169 ACT 1972 57 159792 ACT 1973 35 173306 ACT 1974 34 186241 ACT 1975 67 199007 ACT 1976 44 207740 ACT 1977 145 213688 ACT 1978 180 217981 ACT 1979 280 220797 ACT 1980 269 224291 NSW 1971 3943 4725503 NSW 1972 3698 4795106 NSW 1973 3356 4841898 NSW 1974 3606 4894053 NSW 1975 3517 4932016 NSW 1976 3535 4959588 NSW 1977 3807 5001888 NSW 1978 4180 5053790 NSW 1979 3656 5111130 NSW 1980 3643 5171527 NSW 1981 3841 5234889 NT 1971 412 85735 NT 1972 404 92081 NT 1973 524 97127 NT 1974 563 102924 NT 1975 494 92869 NT 1976 515 98228 NT 1977 560 103938 NT 1978 719 109980 NT 1979 911 114149 NT 1980 722 118245 NT 1981 970 122616 Qld 1971 1852 1851485 Qld 1972 2039 1898478 Qld 1973 2192 1951951 Qld 1974 1952 2008340 Qld 1975 1718 2051362 Qld 1976 1492 2092375 Qld 1977 1678 2129839 Qld 1978 2107 2172047 Qld 1979 1695 2214771 Qld 1980 1838 2265935 Qld 1981 1353 2345208
My code by now is
Code:
twoway (line log_GON year if treated == 1, lcolor(red) xline(1995)) (line log_GON year if treated== 0, lcolor(blue))
What is the more meaningful way of doing it? Should I manually aggregate the data (i.e. summing cases and pop by year and create a new "control" state) or is there a easier way of doing it?
