Overlay separate panel data line plots rather than combine

[Erlend Sunde]

Hi, I have generated two separate panel data line plots in two different figures (by using the xtline function). I want to overlay these two separate line plots rather than combine them. I have not yet found a code that can do this. I have only found an overlay option for twoway graphs. I know that you can overlay many panels in one graph when making a panel data line plot, but the two plots I want to overlay are using two somewhat different datasets, so I think it would be tricky.

I am looking for something like the graph combine code, where you type in the name of the graphs you want to combine.

Thank you.

[Andrew Musau]

See the -overlay- option of xtline.

xtline var1 var2, overlay

I know that you can overlay many panels in one graph when making a panel data line plot, but the two plots I want to overlay are using two somewhat different datasets, so I think it would be tricky.

Are the scales the same? Then append the datasets.

[Erlend Sunde]

Thanks for your reply Andrew Musau. I see that I maybe was a bit unclear regarding the problem. To clarify, my dataset looks like this:

* Example generated by -dataex-. For more info, type help dataex
clear
input float(date TRUST_high countryid) double GDP_percapita
228 1 1 49026.2
229 1 1 49191.9
230 1 1 49660.5
231 1 1 50076.5
232 1 1 50679.1
233 1 1 51495.4
234 1 1 51723.3
235 1 1 51882.7
236 1 1   52021
237 1 1 52665.9
238 1 1 53229.4
239 1 1 53712.4
240 1 1 53975.9
241 1 1 50325.9
242 1 1 52712.8
243 1 1 54728.8
244 1 1 56222.2
245 1 1 57434.6
228 1 2 53649.2
229 1 2 53855.6
230 1 2 54274.4
231 1 2 55088.1
232 1 2 56136.4
233 1 2 56950.7
234 1 2 57103.1
235 1 2 57714.9
236 1 2   57951
237 1 2 57788.3
238 1 2   58109
239 1 2 58406.4
240 1 2 57611.9
241 1 2 51087.8
242 1 2 57265.1
243 1 2   56436
244 1 2 56654.7
245 1 2 59708.1
246 1 2 62434.1
228 1 3 49990.7
229 1 3 50198.8
230 1 3 50398.4
231 1 3 51191.4
232 1 3 51723.7
233 1 3 52406.6
234 1 3 52685.7
235 1 3 53294.4
236 1 3 53349.2
237 1 3 53823.8
238 1 3 54498.9
239 1 3 55411.4
240 1 3   54338
241 1 3 48232.5
242 1 3 54713.8
243 1 3 55049.6
244 1 3 56314.9
245 1 3   58147
246 1 3 60109.1
228 1 4 47831.2
229 1 4 48261.6
230 1 4 48489.5
231 1 4 48947.6
232 1 4 49694.6
233 1 4 50331.3
234 1 4 50331.4
235 1 4 49947.8
236 1 4 49373.4
237 1 4 49532.3
238 1 4 49305.1
239 1 4 49301.3
240 1 4 48290.8
241 1 4   42721
242 1 4 47049.5
243 1 4 48394.3
244 1 4 49554.3
245 1 4 49799.2
246 1 4 51074.4
228 1 5 68623.6
229 1 5 68642.8
230 1 5 69006.8
231 1 5 70062.7
232 1 5 71151.2
233 1 5 72151.5
234 1 5 71736.2
235 1 5   71675
236 1 5 71408.4
237 1 5 71728.1
238 1 5 72184.6
239 1 5 72962.1
240 1 5 72526.9
241 1 5 67972.3
242 1 5 72966.8
243 1 5 73158.3
244 1 5 73638.9
245 1 5 75891.5
246 1 5   78268
228 0 6 23711.3
229 0 6 24299.5
230 0 6 24934.1
231 0 6 25196.3
232 0 6 25546.4
233 0 6 25915.2
end
format %tq date

where TRUST_high indicates if the country is defined as a high-trust or low-trust country.

Then I collapse the dataset and define the panel variable:

collapse (mean) GDP_percapita, by (TRUST_high date)

xtset TRUST_high date

and then I generate the first plot

xtline GDP_percapita, overlay plot1(lc(black) lpattern(dash)) plot2(lc(black)) legend(order(1 "Low-trust countries"2 "High-trust countries"))
name(g1)

which overlays all GDPpercapita line plots for each specific country and generate two lines, one for each panel (high/low trust countries).

and then I generate the second plot, which is fairly similar, but the TRUST_high variable is slightly tweaked in terms of which countries are being defined as high- or low-trust:

xtline GDP_percapita, overlay plot1(lc(black) lpattern(dash)) plot2(lc(black)) legend(order(1 "Low-trust countries"2 "High-trust countries"))
name(g2)

So, the GDP_percapita dataset is the same in both graphs, but the distribution of high- and low-trust countries is what is different between the two graphs. Hence, I do not think the code you sent would work, as it assumes that there are two different dependent variables. Maybe the way around this "issue" is obvious, but I cannot see it as of now.

[Andrew Musau]

The expand command below mimics appending. "new" is an indicator for the appended dataset.

* Example generated by -dataex-. For more info, type help dataex
clear
input float(date TRUST_high countryid) double GDP_percapita
228 1 1 49026.2
229 1 1 49191.9
230 1 1 49660.5
231 1 1 50076.5
232 1 1 50679.1
233 1 1 51495.4
234 1 1 51723.3
235 1 1 51882.7
236 1 1   52021
237 1 1 52665.9
238 1 1 53229.4
239 1 1 53712.4
240 1 1 53975.9
241 1 1 50325.9
242 1 1 52712.8
243 1 1 54728.8
244 1 1 56222.2
245 1 1 57434.6
228 1 2 53649.2
229 1 2 53855.6
230 1 2 54274.4
231 1 2 55088.1
232 1 2 56136.4
233 1 2 56950.7
234 1 2 57103.1
235 1 2 57714.9
236 1 2   57951
237 1 2 57788.3
238 1 2   58109
239 1 2 58406.4
240 1 2 57611.9
241 1 2 51087.8
242 1 2 57265.1
243 1 2   56436
244 1 2 56654.7
245 1 2 59708.1
246 1 2 62434.1
228 1 3 49990.7
229 1 3 50198.8
230 1 3 50398.4
231 1 3 51191.4
232 1 3 51723.7
233 1 3 52406.6
234 1 3 52685.7
235 1 3 53294.4
236 1 3 53349.2
237 1 3 53823.8
238 1 3 54498.9
239 1 3 55411.4
240 1 3   54338
241 1 3 48232.5
242 1 3 54713.8
243 1 3 55049.6
244 1 3 56314.9
245 1 3   58147
246 1 3 60109.1
228 1 4 47831.2
229 1 4 48261.6
230 1 4 48489.5
231 1 4 48947.6
232 1 4 49694.6
233 1 4 50331.3
234 1 4 50331.4
235 1 4 49947.8
236 1 4 49373.4
237 1 4 49532.3
238 1 4 49305.1
239 1 4 49301.3
240 1 4 48290.8
241 1 4   42721
242 1 4 47049.5
243 1 4 48394.3
244 1 4 49554.3
245 1 4 49799.2
246 1 4 51074.4
228 1 5 68623.6
229 1 5 68642.8
230 1 5 69006.8
231 1 5 70062.7
232 1 5 71151.2
233 1 5 72151.5
234 1 5 71736.2
235 1 5   71675
236 1 5 71408.4
237 1 5 71728.1
238 1 5 72184.6
239 1 5 72962.1
240 1 5 72526.9
241 1 5 67972.3
242 1 5 72966.8
243 1 5 73158.3
244 1 5 73638.9
245 1 5 75891.5
246 1 5   78268
228 0 6 23711.3
229 0 6 24299.5
230 0 6 24934.1
231 0 6 25196.3
232 0 6 25546.4
233 0 6 25915.2
end
format %tq date

expand 2, g(new)
set seed 05312022
replace TRUST_high= runiformint(0,1) if new
egen group= group(TRUST_high new)
collapse GDP_percapita, by (group date)
xtset group date
xtline GDP, overlay

[Erlend Sunde]

Thank you Andrew Musau. It seems like that code just generates a new TRUST_high variable, but it does not show how to overlay the graph using the first TRUST_high variable with the other (tweaked) TRUST_high variable (lets call it TRUST_high2). Essentially, what I have is two different panel variables (TRUST_high and TRUST_high2). I have made two different separate plots using each of these panel variables, and I want to overlay these two plots.

To clarify I can upload the graphs here. My first plot, using the panel variable TRUST_high, looks like this:

wwhile the second plot, using the panel variable TRUST_high2, looks similar to the graph above, but with some tweaks in levels, trends etc.

[Andrew Musau]

We are going round in circles here. The idea is to create a categorical variable with 4 levels in the appended dataset, and this will give you what you want. Provide a data example including TRUST_high and TRUST_high2 if you are not able to implement the advice in #4.

[Erlend Sunde]

It worked! Sorry, I did not quite understand the code at first, but eventually it made sense. Thank you Andrew Musau